Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018: Selected Papers from the ICOSAHOM Conference, London, UK, July 9-13, 2018 [1st ed.] 9783030396466, 9783030396473

Citation preview

134

Spencer J. Sherwin · David Moxey Joaquim Peiró · Peter E. Vincent Christoph Schwab Editors

Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018

Editorial Board T. J.Barth M.Griebel D.E.Keyes R.M.Nieminen D.Roose T.Schlick

Lecture Notes in Computational Science and Engineering Editors: Timothy J. Barth Michael Griebel David E. Keyes Risto M. Nieminen Dirk Roose Tamar Schlick

134

More information about this series at http://www.springer.com/series/3527

Spencer J. Sherwin • David Moxey • Joaquim Peiró • Peter E. Vincent • Christoph Schwab Editors

Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018 Selected Papers from the ICOSAHOM Conference, London, UK, July 9–13, 2018

Editors Spencer J. Sherwin Department of Aeronautics Imperial College London, UK

Joaquim Peiró Department of Aeronautics Imperial College London, UK

David Moxey College of Engineering, Mathematics & Physical Sciences University of Exeter Exeter, UK Peter E. Vincent Department of Aeronautics Imperial College London, UK

Christoph Schwab Department of Mathematics ETH Zürich Zürich, Switzerland

ISSN 1439-7358 ISSN 2197-7100 (electronic) Lecture Notes in Computational Science and Engineering ISBN 978-3-030-39646-6 ISBN 978-3-030-39647-3 (eBook) https://doi.org/10.1007/978-3-030-39647-3 Mathematics Subject Classification: 65M70, 65N35, 65N30, 74S25, 76M10, 76M22, 78M10, 78M22 This book is an open access publication. © The Editor(s) (if applicable) and The Author(s) 2020 Open Access This book is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this book are included in the book’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the book’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Cover illustration: A 7th order accurate simulation of free stream turbulence passing over a turbine blade simulated using the Nektar++ package, courtesy of Andrea Cassinelli This Springer imprint is published by the registered company Springer Nature Switzerland AG. The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

This volume presents selected papers from the twelfth International Conference on Spectral and High-Order Methods (ICOSAHOM’18) that was held in London, United Kingdom, during the week of July 9–13th, 2018. These selected papers were refereed by members of the scientific committee of ICOSAHOM, as well as by other leading scientists. The first ICOSAHOM conference was held in Como, Italy, in 1989 and marked the beginning of an international conference series in Montpellier, France (1992); Houston, TX, USA (1995); Tel Aviv, Israel (1998); Uppsala, Sweden (2001); Providence, RI, USA (2004); Beijing, China (2007); Trondheim, Norway (2009); Gammarth, Tunisia (2012); Salt Lake City, USA (2014); and Rio de Janeiro, Brazil (2016). ICOSAHOM has established itself as the main meeting place for researchers with interests in the theoretical, applied, and computational aspects of high-order methods for the numerical solution of partial differential equations. With over 360 attendees, ICOSAHOM ’18 has been the largest edition of the conference series to date. The program consisted of eight invited speakers across the week from internationally renowned researchers, alongside 40 minisymposia (of around 300 presentations) dedicated to specialized topics in high-order methods, and approximately a further 90 contributed talks. The content of these proceedings is organized as follows. First, contributions from the invited speakers are included. The remainder of the volume consists of refereed selected papers highlighting the broad spectrum of topics presented at ICOSAHOM ’18. The success of ICOSAHOM ’18 was ensured through generous contributions and financial support of our sponsors: the Air Force Office of Scientific Research (AFSOR); the Platform for Research in Simulation Methods (PRISM) platform grant, funded by the Engineering and Physical Sciences Research Council (EPSRC); Rolls-Royce Ltd.; and, finally, the Department of Aeronautics at Imperial College London. We would like to give special thanks to our local organizing committee for their efforts in organizing and promoting the event. In particular, we would also v

vi

Preface

like to thank Mr. Andrea Cassinelli for his organizational efforts leading up to the conference, as well as the administrative staff of the Department of Aeronautics at Imperial College London for their help in coordinating the logistics of the event. We also thank the many student helpers for their advice, help, and support given to the delegates during the event itself, who all contributed to the smooth running of the event. London, UK Exeter, UK London, UK London, UK Zürich, Switzerland

Spencer J. Sherwin David Moxey Joaquim Peiró Peter E. Vincent Christoph Schwab

Contents

Part I

Invited Papers

Stability of Wall Boundary Condition Procedures for Discontinuous Galerkin Spectral Element Approximations of the Compressible Euler Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Florian J. Hindenlang, Gregor J. Gassner, and David A. Kopriva

3

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Florian J. Hindenlang and Gregor J. Gassner

21

A Review of Regular Decompositions of Vector Fields: Continuous, Discrete, and Structure-Preserving . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Ralf Hiptmair and Clemens Pechstein

45

Model Reduction by Separation of Variables: A Comparison Between Hierarchical Model Reduction and Proper Generalized Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Simona Perotto, Michele Giuliano Carlino, and Francesco Ballarin Recurrence Relations for a Family of Orthogonal Polynomials on a Triangle .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Sheehan Olver, Alex Townsend, and Geoffrey M. Vasil Part II

61

79

Contributed Papers

Greedy Kernel Methods for Center Manifold Approximation .. . . . . . . . . . . . . Bernard Haasdonk, Boumediene Hamzi, Gabriele Santin, and Dominik Wittwar

95

An Adaptive Error Inhibiting Block One-Step Method for Ordinary Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 107 Jiaxi Gu and Jae-Hun Jung

vii

viii

Contents

Hermite Methods in Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 119 Rujie Gu and Thomas Hagstrom HPS Accelerated Spectral Solvers for Time Dependent Problems: Part II, Numerical Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 131 Tracy Babb, Per-Gunnar Martinsson, and Daniel Appelö On the Use of Hermite Functions for the Vlasov–Poisson System .. . . . . . . . . 143 Lorella Fatone, Daniele Funaro, and Gianmarco Manzini HPS Accelerated Spectral Solvers for Time Dependent Problems: Part I, Algorithms .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 155 Tracy Babb, Per-Gunnar Martinsson, and Daniel Appelö High-Order Finite Element Methods for Interface Problems: Theory and Implementations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 167 Yuanming Xiao, Fangman Zhai, Linbo Zhang, and Weiying Zheng Stabilised Hybrid Discontinuous Galerkin Methods for the Stokes Problem with Non-standard Boundary Conditions . . . . . .. . . . . . . . . . . . . . . . . . . . 179 Gabriel R. Barrenechea, Michał Bosy, and Victorita Dolean RBF Based CWENO Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 191 Jan S. Hesthaven, Fabian Mönkeberg, and Sara Zaninelli Discrete Equivalence of Adjoint Neumann–Dirichlet div-grad and grad-div Equations in Curvilinear 3D Domains . . . . .. . . . . . . . . . . . . . . . . . . . 203 Yi Zhang, Varun Jain, Artur Palha, and Marc Gerritsma A Conservative Hybrid Method for Darcy Flow. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 215 Varun Jain, Joël Fisser, Artur Palha, and Marc Gerritsma High-Order Mesh Generation Based on Optimal Affine Combinations of Nodal Positions .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 229 Mike Stees and Suzanne M. Shontz Sparse Spectral-Element Methods for the Helically Reduced Einstein Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 239 Stephen R. Lau Spectral Analysis of Isogeometric Discretizations of 2D Curl-Div Problems with General Geometry .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 251 Mariarosa Mazza, Carla Manni, and Hendrik Speleers Performance of Preconditioners for Large-Scale Simulations Using Nek5000 . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 263 N. Offermans, A. Peplinski, O. Marin, E. Merzari, and P. Schlatter Two Decades Old Entropy Stable Method for the Euler Equations Revisited .. . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 273 Björn Sjögreen and H. C. Yee

Contents

ix

A Mimetic Spectral Element Method for Free Surface Flows . . . . . . . . . . . . . . 285 L. Nielsen and B. Gervang Spectral/hp Methodology Study for iLES-SVV on an Ahmed Body . . . . . . . 297 Filipe F. Buscariolo, Spencer J. Sherwin, Gustavo R. S. Assi, and Julio R. Meneghini A High-Order Discontinuous Galerkin Solver for Multiphase Flows .. . . . . 313 Juan Manzanero, Carlos Redondo, Gonzalo Rubio, Esteban Ferrer, Eusebio Valero, Susana Gómez-Álvarez, and Ángel Rivero-Jiménez High-Order Propagation of Jet Noise on a Tetrahedral Mesh Using Large Eddy Simulation Sources .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 325 M. A. Moratilla-Vega, V. Saini, H. Xia, and G. J. Page Dynamical Degree Adaptivity for DG-LES Models . . . . . .. . . . . . . . . . . . . . . . . . . . 337 M. Tugnoli, A. Abbà, and L. Bonaventura A Novel Eighth-Order Diffusive Scheme for Unstructured Polyhedral Grids Using the Weighted Least-Squares Method .. . . . . . . . . . . . . 349 Duarte M. S. Albuquerque, Artur G. R. Vasconcelos, and Jose C. F. Pereira An Explicit Mapped Tent Pitching Scheme for Maxwell Equations. . . . . . . . 359 Jay Gopalakrishnan, Matthias Hochsteger, Joachim Schöberl, and Christoph Wintersteiger Viscous Diffusion Effects in the Eigenanalysis of (Hybridisable) DG Methods . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 371 Rodrigo C. Moura, Pablo Fernandez, Gianmarco Mengaldo, and Spencer J. Sherwin Spectral Galerkin Method for Solving Helmholtz and Laplace Dirichlet Problems on Multiple Open Arcs . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 383 Carlos Jerez-Hanckes and José Pinto Explicit Polynomial Trefftz-DG Method for Space-Time Elasto-Acoustics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 395 H. Barucq, H. Calandra, J. Diaz, and E. Shishenina An hp-Adaptive Iterative Linearization Discontinuous-Galerkin FEM for Quasilinear Elliptic Boundary Value Problems.. . . . . . . . . . . . . . . . . . . 407 Paul Houston and Thomas P. Wihler Erosion Wear Evaluation Using Nektar++ .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 419 Manuel F. Mejía, Douglas Serson, Rodrigo C. Moura, Bruno S. Carmo, Jorge Escobar-Vargas, and Andrés González-Mancera An Inexact Petrov-Galerkin Approximation for Gas Transport in Pipeline Networks .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 429 Herbert Egger, Thomas Kugler, and Vsevolod Shashkov

x

Contents

New Preconditioners for Semi-linear PDE-Constrained Optimal Control in Annular Geometries . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 441 Lasse Hjuler Christiansen and John Bagterp Jørgensen DIRK Schemes with High Weak Stage Order . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 453 David I. Ketcheson, Benjamin Seibold, David Shirokoff, and Dong Zhou Scheme for Evolutionary Navier-Stokes-Fourier System with Temperature Dependent Material Properties Based on Spectral/hp Elements. . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 465 Jan Pech Implicit Large Eddy Simulations for NACA0012 Airfoils Using Compressible and Incompressible Discontinuous Galerkin Solvers.. . . . . . . 477 Esteban Ferrer, Juan Manzanero, Andres M. Rueda-Ramirez, Gonzalo Rubio, and Eusebio Valero SAV Method Applied to Fractional Allen-Cahn Equation . . . . . . . . . . . . . . . . . . 489 Xiaolan Zhou, Mejdi Azaiez, and Chuanju Xu A First Meshless Approach to Simulation of the Elastic Behaviour of the Diaphragm.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 501 Nicola Cacciani, Elisabeth Larsson, Alberto Lauro, Marco Meggiolaro, Alessio Scatto, Igor Tominec, and Pierre-Frédéric Villard An Explicit Hybridizable Discontinuous Galerkin Method for the 3D Time-Domain Maxwell Equations . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 513 Georges Nehmetallah, Stéphane Lanteri, Stéphane Descombes, and Alexandra Christophe Entropy Conserving and Kinetic Energy Preserving Numerical Methods for the Euler Equations Using Summation-by-Parts Operators . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 525 Hendrik Ranocha Multiwavelet Troubled-Cell Indication: A Comparison of Utilizing Theory Versus Outlier Detection . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 537 Mathea J. Vuik An Anisotropic p-Adaptation Multigrid Scheme for Discontinuous Galerkin Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 549 Andrés M. Rueda-Ramírez, Gonzalo Rubio, Esteban Ferrer, and Eusebio Valero A Spectral Element Reduced Basis Method for Navier–Stokes Equations with Geometric Variations.. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 561 Martin W. Hess, Annalisa Quaini, and Gianluigi Rozza

Contents

xi

Iterative Spectral Mollification and Conjugation for Successive Edge Detection.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 573 Robert E. Tuzun and Jae-Hun Jung Small Trees for High Order Whitney Elements . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 587 Ana Alonso Rodríguez and Francesca Rapetti Non-conforming Elements in Nek5000: Pressure Preconditioning and Parallel Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 599 A. Peplinski, N. Offermans, P. F. Fischer, and P. Schlatter Sparse Approximation of Multivariate Functions from Small Datasets Via Weighted Orthogonal Matching Pursuit . . .. . . . . . . . . . . . . . . . . . . . 611 Ben Adcock and Simone Brugiapaglia On the Convergence Rate of Hermite-Fejér Interpolation . . . . . . . . . . . . . . . . . . 623 Shuhuang Xiang and Guo He Fifth-Order Finite-Volume WENO on Cylindrical Grids . . . . . . . . . . . . . . . . . . . 637 Mohammad Afzal Shadab, Xing Ji, and Kun Xu

Part I

Invited Papers

Stability of Wall Boundary Condition Procedures for Discontinuous Galerkin Spectral Element Approximations of the Compressible Euler Equations Florian J. Hindenlang, Gregor J. Gassner, and David A. Kopriva

1 Introduction The ingredients for a reliable numerical method for the approximation of partial differential equations, e.g. one that will not blow up, include stable inter-element and physical boundary condition implementations. The recognition that the discontinuous Galerkin spectral element method (DGSEM) with Gauss-Lobatto quadratures satisfies a summation-by-parts (SBP) operators [4, 7] has allowed for the analysis of these schemes and to connect them with penalty collocation and SBP finite difference schemes. For instance, in [5], we showed that a split form approximation of the compressible Navier–Stokes equations was both linearly and entropy stable provided that the boundary conditions were properly imposed. The importance of stable boundary condition procedures for hyperbolic equations has long been studied, especially in relation to finite difference methods, e.g. [3, 9, 10]. Only recently have they been studied for discontinuous Galerkin approximations. In [12], the authors showed that the reflection approach is stable when using an entropy conserving flux and an additional entropy stable dissipation

F. J. Hindenlang Max Planck Institute for Plasma Physics, Garching, Germany e-mail: [email protected] G. J. Gassner () Department for Mathematics and Computer Science, Center for Data and Simulation Science, University of Cologne, Cologne, Germany e-mail: [email protected] D. A. Kopriva Florida State University, Tallahassee, FL, USA San Diego State University, San Diego, CA, USA e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_1

3

4

F. J. Hindenlang et al.

term (EC-ES). In [2], the authors show that the reflection condition is stable if the numerical flux is either the Godunov or HLL flux. In this paper, we analyze both the linear and entropy stability of two types of commonly used wall boundary condition procedures used with the DGSEM applied to the compressible Euler equations. In both cases, wall boundary conditions are implemented through a numerical flux. The boundary condition might be implemented through a special wall numerical flux that includes the boundary condition, or a fictitious external state applied to a Riemann solver approximation. We show how to construct special wall numerical fluxes that are stable, and study the behavior of the approximations. In particular, we show that the use of Riemann solvers at the boundaries introduce numerical dissipation in an amount that depends on the size of the normal Mach number at the wall.

2 The Compressible Euler Equations and the Wall Boundary Condition We write the Euler equations as ut +

3  ∂fi = 0. ∂xi

(1)

i=1

The state vector contains the conservative variables  T  T u =   v E =  v1 v2 v3 E .

(2)

In standard form, the components of the advective fluxes are ⎤ v1 ⎢ v 2 + p ⎥ ⎥ ⎢ 1 ⎥ ⎢ f1 = ⎢ v1 v2 ⎥ ⎥ ⎢ ⎣ v1 v3 ⎦ (E + p)v1 ⎡

⎤ v2 ⎢ v v ⎥ ⎢ 2 1 ⎥ ⎥ ⎢ f2 = ⎢ v22 + p ⎥ ⎥ ⎢ ⎣ v2 v3 ⎦ (E + p)v2 ⎡

⎤ v3 ⎢ v v ⎥ ⎢ 3 1 ⎥ ⎥ ⎢ f3 = ⎢ v3 v2 ⎥ , ⎥ ⎢ ⎣ v32 + p ⎦ (E + p)v3 ⎡

(3)

Here, , v = (v1 , v2 , v3 )T , p, E are the mass density, fluid velocities, pressure and total energy. We close the system with the ideal gas assumption, which relates the total energy and pressure 

1 2 p = (γ − 1) E −  v , 2

(4)

Stability of Wall Boundary Condition Procedures

5

where γ denotes the adiabatic coefficient. For a compact notation that simplifies the analysis, we define block vectors (with the double arrow)  T f = f1 f2 f3 ,

↔

(5)

so that the system of equations can be written in the compact form ↔

 x · f = 0. ut + ∇

(6)

The linear Euler equations are derived by linearizing about a constant mean state (, ¯ v¯1 , v¯2 , v¯3 , p). ¯ We follow [11] for the symmetrization of the linearized equations, with the constants   c¯ γ −1 γ p¯ c, ¯ b = √ , c¯ = , (7) a= γ γ ¯ where c¯ is the sound speed of the constant mean state. The state variables become  T u =  v1 v2 v3 p ,

(8)

where v is the velocity perturbation from the mean state, and we introduce ˜  = b , ¯

p =

1 1 p˜ − √  , a ¯ γ −1

(9)

which depend on the density and pressure perturbations , ˜ p. ˜ The flux vectors are fi = Ai u,

   = A1 xˆ + A2 yˆ + A3 zˆ u, f = Au

↔

(10)

where [11] ⎡

v¯1 ⎢b ⎢ ⎢ A1 = ⎢ 0 ⎢ ⎣0 0

b v¯1 0 0 a

0 0 v¯1 0 0

0 0 0 v¯1 0

⎤ 0 a⎥ ⎥ ⎥ 0 ⎥, ⎥ 0⎦ v¯1

⎡

v¯2 ⎢0 ⎢ ⎢ A2 = ⎢ b ⎢ ⎣0 0

are constant symmetric matrices.

0 v¯2 0 0 0

b 0 v¯2 0 a

0 0 0 v¯2 0

⎤ 0 0⎥ ⎥ ⎥ a ⎥, ⎥ 0⎦ v¯2

⎡

v¯3 ⎢0 ⎢ ⎢ A3 = ⎢ 0 ⎢ ⎣b 0

0 v¯3 0 0 0

0 0 v¯3 0 0

b 0 0 v¯3 a

⎤ 0 0⎥ ⎥ ⎥ 0⎥ ⎥ a⎦ v¯3 (11)

6

F. J. Hindenlang et al.

The linear equations have the property that the L2 norm of the solution over a domain  is bounded by terms of the boundary data on ∂, only. Let  v, w =

T

v w dxdydz,

3 ↔    ↔ f, g = fTi gi dxdydz.

(12)

 i=1



↔

represent the L2 inner product of two state vectors v and w and two block vectors f ↔ and g, respectively. Since the coefficient matrices are constant the product rule and  implies symmetry of A         ↔ ↔  x · f, u = ∇  x · Au  , u = ∇x u, f . ∇

(13)

Then it follows from Gauss’ law (integration by parts) that 

 1 ↔ ↔  ∇x · f, u = uT f · ndS, 2 ∂

(14)

where n is the outward normal to the surface of . The norm of the solution therefore satisfies  ↔ d 2 ||u|| = − uT f · n dS. (15) dt ∂ Replacing the boundary terms by boundary conditions leads to a bound on the solution in terms of the boundary data. The argument of the boundary integral on the right of (15) is     ↔  · n u = 2 b + ap vn + (v¯ · n )(2 + | uT f · n = uT A v |2 + p2 ),

(16)

where vn is the wall normal velocity, vn = v · n . Note that here, the mean flow must be chosen such that the normal flow vanishes at the wall boundary v¯ · n = 0, so that the boundary condition makes physical sense. Therefore, with the no penetration wall condition vn = 0 applied, d ||u||2 = 0, dt

(17)

and the (energy) norm of the solution is bounded for all time by its initial value. The nonlinear equations, on the other hand, satisfy a bound on the entropy that depends only on the boundary data. For what follows, we assume that the solution is smooth so that we don’t have to consider entropy generated at shock waves. We

Stability of Wall Boundary Condition Procedures

7

introduce the entropy density (scaled with (γ − 1) for convenience) as s(u) = −

ς , (γ − 1)

(18)

where ς = ln(p) − γ ln() is the physical entropy. (The minus sign is conventional in the theory of hyperbolic conservation laws to ensure a decreasing entropy function.) The entropy flux for the Euler equations is ς v f ς (u) = v s = − . (γ − 1)

(19)

Finally the entropy variables are ⎡ γ −ς w=

∂s(u) ⎢ =⎣ ∂u

γ −1

⎤ − β|| v ||2 , ⎥ ⎦, 2β v −2β

β=

 . 2p

(20)

The entropy pair contracts the solution and fluxes, meaning that it satisfies the relations  w ut = T

∂s ∂u

T ut = st (u),

↔

 x · f ς . x · f = ∇ wT ∇

(21)

When we multiply (6) with the entropy variables and integrate over the domain, 

   ↔ x · f = 0 . w(u), ut + w(u), ∇

(22)

Next we use the properties of the entropy pair to contract (22) and use integration by parts to get      x · f ς , 1 = − st (u), 1 = − ∇

   f ς · n dS

(23)

∂

showing that, in the continuous case, the total entropy in the domain can only change via the boundary conditions. In the case of a zero-mass flux boundary condition, with vn = v · n = 0, the entropy is not changed by the slip-wall boundary condition, since − f ς · n =

ς vn = 0. (γ − 1)

(24)

8

F. J. Hindenlang et al.

3 Stability Bounds for the DGSEM The DGSEM is described in detail in [5] and elsewhere [1, 6]. We will only quickly summarize the approximation here. The domain,  is subdivided into non-overlapping, conforming, hexahedral elements. Each element is mapped to the reference element E = [−1, 1]3. Associated with the transformation from the reference element is a set of contravariant coordinate vectors, a i , and transformation Jacobian, J. Equation (6) transform to another conservation law on the reference element as ↔

 ξ · ˜f = 0, Jut + ∇ ↔

(25) ↔

where ˜f is the contravariant flux vector with components ˜fi = Ja i · f. The approximation of (25) proceeds as follows: A weak form is created by taking the inner product of the equation with a test function. The Gauss law is applied to the divergence term to separate the boundary from the interior contributions. The resulting weak form is then approximated: The solution vector is approximated by a polynomial of degree N interpolated at the Legendre–Gauss–Lobatto points. In the following, we will represent the true continuous solutions by lower case letter. Upper case letters will denote their polynomial approximations, except for the density, where the approximation is denoted by ρ. The volume fluxes are replaced by twopoint numerical fluxes. In the linear case, the two point fluxes are immediately relatable to a split form of the equations. Integrals are replaced by Legendre–Gauss– Lobatto quadratures. Finally, the boundary fluxes are replaced by a numerical flux. See [5] and [8] for details. The result is an approximation that is energy stable for the linearized equations if at every quadrature point along a physical boundary the numerical flux F˜ ∗ satisfies the bound [5]   1↔ UT F˜ ∗ − F˜ · nˆ ≥ 0, 2

(26)

↔

where F˜ is the polynomial interpolation of the contravariant flux from the interior, nˆ is the reference space outward normal direction, and U is the approximation of the state vector. Since the contravariant fluxes are proportional to the normal fluxes [6], we can change the condition (26) to   1↔ BL ≡ UT F∗ − F · n ≥ 0, 2

(27)

Stability of Wall Boundary Condition Procedures

9

For entropy stability of the nonlinear equations, the boundary stability condition shown in [5] is proportional to BNL ≡ WT

  ↔    F∗ − F · n + F ς · n ≥ 0,

(28)

where F ς is the polynomial interpolation of the entropy flux, f ς , and W is the interpolation of the entropy variables.

3.1 Linear Stability of Wall Boundary Condition Approximations To find linearly stable implementations of the wall condition vn = 0, one needs only find a numerical flux that satisfies it and the condition (27). For the linear equations, the approximation of the state vector is U = [ρ  V P  ]T and the normal contravariant flux is proportional to   ↔  · n U = bVn n1 Q n2 Q n3 Q aVn T , F · n = A

(29)

where Vn is the approximation of the normal velocity at the wall computed from the interior, Q = bρ  + aP  , and (n1 , n1 , n3 ) are the three components of the physical space normal vector, n. The numerical flux can be expressed as    · n U∗ = bVn∗ n1 Q∗ n2 Q∗ n3 Q∗ aVn∗ T . F∗ = A

(30)

It then remains only to find Q∗ so that (27) is satisfied when the normal wall condition Vn∗ = 0 is applied. When we substitute the fluxes (29) and (30) into (27), BL =

    1   1  ∗ Q 2Vn − Vn + Vn 2Q∗ − Q = 2QVn∗ + 2Vn Q∗ − Q 2 2 (31)

Substituting the wall boundary condition Vn∗ = 0 yields the condition on Q∗ for stability   Vn Q∗ − Q ≥ 0.

(32)

Neutral stability is thus ensured if ρ ∗ and P ∗ are computed from the interior, i.e. ρ ∗ = ρ  , P ∗ = P  so that Q∗ = Q. In practice, the boundary condition is also implemented through the use of a Riemann solver and external state designed to imply the physical boundary

10

F. J. Hindenlang et al.

condition to construct the numerical boundary flux. The exact upwind (ε = 1) normal Riemann flux and the central flux (ε = 0) for the linear system of equations is   1 ↔   ↔   ε    F∗ U, Uext = F U · n + F Uext · n − A n  Uext − U , 2 2

(33)

 · n is the normal coefficient matrix. The external state is set by using where A n ≡ A the interior values of the density and pressure and the negative of the value of the normal velocity, U

ext

 = ρ

  V − 2Vn n P 

T

.

(34)

For ε = 0, using the central (averaged) numerical flux, the interior flux contribution cancels and condition (27) reduces to ⎡

BL,0

0 ⎢n b ! "⎢ 1 1 ⎢ = UT A n Uext = ρ  V P  ⎢ n2 b ⎢ 2 ⎣ n3 b 0     = Q −V · n + V · n Q = 0,

n1 b 0 0 0 n1 a

n2 b 0 0 0 n2 a

n3 b 0 0 0 n3 a

⎤ ⎤⎡ ρ 0 ⎢ ⎥ n1 a ⎥ ⎥ ⎥⎢ ⎥ ⎥⎢  n2 a ⎥ ⎢ V − 2Vn n ⎥ ⎥ ⎥⎢ ⎦ n3 a ⎦ ⎣ P 0 (35)

which is neutrally stable, having no additional stabilizing dissipation. We note again, that the mean state for the linearization is chosen such that the normal mean velocity components are zero, resulting in the zeros on the diagonal of A n . Substituting the exact upwind flux where ε = 1 into (27) and rearranging,     1   (36) BL,1 = −UT A−n  Uext + UT A n  U, 2    where A−n = 12 A n − A n  is negative semidefinite. The second term is nonnegative, depends only on the interior state, and adds stabilizing dissipation. From the matrix absolute value, the dissipation term is   1 UT A n  U = Q2 + c¯3 Ma2n , c¯

(37)

where Man = Vn /c¯ is the normal Mach number. Stability depends, then, on the value of the first term, which is where the boundary conditions are incorporated

Stability of Wall Boundary Condition Procedures

11

through the external state Uext written in (34). Then   1 c¯3   UT A−n  Uext = Q2 − Ma2 . 2c¯ 2

(38)

Therefore, using the upwind numerical flux, (36) becomes BL,1 = c¯3 Ma2n ≥ 0,

(39)

as required. The amount of dissipation depends on how far the interior computed normal velocity deviates from zero. The combination of the reflective state and local Lax-Friedrichs flux is also linearly stable.  In that case the exact matrix absolute value is replaced by a diagonal matrix, An  ≈ |λ|max I. The jump term is added to the central (averaged) flux so BL,LF = −

 |λ|max T  ext U U − U = c¯2 |λ|max Ma2n ≥ 0 2

(40)

Finally, a dissipative version of the direct numerical flux (30) can be formed by looking at the reflective state approach. For instance, the equivalent to using the Lax-Friedrichs flux is to choose ρ ∗ = ρ  and P∗ = P +

c¯3 |λ| Man . a max

(41)

Then Q∗ = Q + c¯3 |λ|max Man and   Vn Q∗ − Q = c¯2 |λ|max Ma2n ≥ 0.

(42)

A similar, though more complicated, modified P ∗ can be made to be equivalent to the exact upwind flux.

3.2 Entropy Stability of Wall Boundary Condition Approximations As in the linear approximation, the wall boundary condition can be imposed for the nonlinear equations either by directly specifying the numerical flux or by computing it through a Riemann solver using a reflection external state that enforces the normal wall condition implicitly. Note that in this section, the discrete variables (ρ, V , P ) describe the full nonlinear state.

12

F. J. Hindenlang et al.

For the nonlinear equations, we construct the numerical flux for a slip-wall as ⎡ ⎤ 0  ↔ ∗ ⎢ ⎥ F · n = ⎣ P ∗ n ⎦ 0

(43)

where we imposed Vn = 0 leading to a flux with no mass or energy transfer, and we introduce a wall pressure P ∗ , whose value will be chosen to ensure consistency and stability. After some manipulations, the discrete entropy stability condition (28) becomes  −ρVn

γ −ς − β||V ||2 (γ − 1)

    ρςVn +2βVn P ∗ − P − ρ||V ||2 + 2βVn ρE + P − = (γ − 1) 

  γ − β||V ||2 + 2βVn P ∗ + ρE − ρ||V ||2 = −ρVn (γ − 1) 

γ ρVn P 1 1 − P + ρ||V ||2 + P ∗ + − ρ||V ||2 = P (γ − 1) 2 (γ − 1) 2 $ # P∗ ρVn −1 ≥0 P

(44)

Therefore if we choose P ∗ = P , to be the internal pressure, the boundary flux does not contribute to the total entropy, independent of the inner normal velocity Vn . A value of P ∗ that leads to a dissipative boundary condition can be found either through exact solution of the Riemann problem at the boundary, or through the use of an external state and an approximate Riemann solver.

3.2.1 Exact Solution of the Riemann Problem In [14] a symmetric 1D Riemann problem is exactly solved following Toro [13], to get the wall pressure P ∗ , accounting for the fact that Vn never vanishes discretely and therefore the wall pressure should be different from the interior pressure. The exact solution of the 1D Riemann problem reads as #

P∗ P

$ = RP

⎧ $ # ) 2 ⎪ ⎪ (γ +1) (γ +1) ⎪ 1 + γ Ma Ma + Ma + 1 >1 ⎨ n n n 4 4

for

Vn > 0

⎪   2γ ⎪ ⎪ ⎩ 1 + 1 (γ − 1)Man (γ −1) 2

for

Vn ≤ 0

≤1

(45) with the normal Mach number, Man =

Vn c ,

and the sound speed c =

*

γ Pρ .

Stability of Wall Boundary Condition Procedures

13

As shown by Toro [13], the solution for the rarefaction has a limiting vacuum solution for Man ≤ −2(γ − 1)−1 . We will restrict our analysis to normal Mach numbers yielding strictly positive pressure solutions only (Man > −5 for γ = 75 ). It is easy to see that using P ∗ from (45), the entropy inequality (44) is still satisfied for |Vn | = 0, and the added entropy scales with the discrete value of Vn at the boundary. Hence, for h → 0, the discrete boundary condition converges to its physical counterpart, since Vn → 0. The choice of P ∗ from (45) appears to stabilize under-resolved simulations, which can be now explained by the fact that the boundary flux always adds entropy for |Vn | = 0.

3.2.2 Using Approximate Riemann Solvers for the Boundary Flux A well known strategy in finite volume methods is to mirror only the velocity of the internal state and solve an approximate Riemann problem to get the boundary flux, mostly just because of a simpler implementation, since an approximate Riemann solver is already available and used for the fluxes between the elements. For DG methods, see also, for example, [2] and [12] where reflection conditions are proved to be entropy stable. The mirror state is set so that the mass and energy flux are zero. Let the inner state be labeled L and the outer R. then the inner and outer states that satisfy the mirror condition are ! "T UL = ρ ρ Vn E ,

! "T UR = ρ ρ(V − 2Vn n) E

(46)

We show below under what conditions on the normal velocity Vn that the reflection condition is entropy stable for the Lax-Friedrichs, HLL and HLLC, Roe and EC-ES fluxes.

Lax-Friedrichs Flux We start with the simplest approximate Riemann solver, the Lax-Friedrichs or Rusanov flux, which reads as  ↔ ∗  |λ| ↔ 1 ↔ F · n = n · F(UL ) + F(UR ) − max (UR − UL ). LF 2 2

(47)

Inserting the states from (46), we get ⎡ ⎤ 0 ↔ ∗ ⎢ ⎥ λmax F · n =⎢ (ρVn2 + P ) n ⎥ ⎣ ⎦− 2 LF 0

⎤ ⎡ ⎤ 0 0 ⎥ ⎢ ⎥ ⎢ ⎢ −2ρVn n ⎥ = ⎢ (ρV 2 + ρVn λmax + P ) n ⎥ . n ⎦ ⎣ ⎦ ⎣ 0 0 ⎡

(48)

14

F. J. Hindenlang et al.

The maximum wave speed is normally approximated from the largest leftgoing and rightgoing wave speed, λmax = max(|VnL | + cL , |VnR | + cR ) = |Vn | + c ,

since cL = cR = c ,

Vn = VnL = −VnR

(49)

and thus gives a definition of P ∗ 

P∗ P

  = 1 + γ Man Man + |Man | + 1 LF

+ =

1 + γ Man (2Man + 1) > 1 ≤1 1 + γ Man

for Vn > 0 , for Vn ≤ 0

(50)

which shows that the Lax-Friedrichs flux satisfies the entropy inequality (44).

HLL and HLLC Flux The HLL flux [13] is written as  ↔ ∗ F · n

HLL

1 = R S − SL

     ↔ ↔ R L L R L R R L n · S F(U ) − S F(U ) + S S U − U . (51)

The leftgoing and rightgoing wave speeds are S L = VnL − cL = −VnR − cR = −S R and the HLL flux reduces to  ↔ ∗ F · n

HLL

  SR  ↔ 1 ↔ L n · F(U ) + F(UR ) − UR − UL . 2 2

=

(52)

If we would choose S R to be the maximum wave speed, the HLL flux would reduce to the Lax-Friedrichs flux. However, with S R = VnR + cR = −Vn + c, an even simpler relation for P ∗ is found, which also satisfies the entropy inequality 

P∗ P

+

= 1 + γ Man HLL

>1 ≤1

for Vn > 0 for Vn ≤ 0

(53)

For the HLLC flux [13], one can show that since the Riemann problem is symmetric, the approximate wave speed of the contact discontinuity is λ∗ = 0 and, choosing S R = −Vn + c, HLLC reduces to the HLL flux. 

P∗ P



= 1 + γ Man = HLLC

P∗ P

(54) HLL

Stability of Wall Boundary Condition Procedures

15

Roe Flux For the original Roe method without entropy fix [13], the mean values are , ρ L VnL + ρ R VnR , , = 0, ρL + ρR ) (γ − 1) 2 Man . c˜ = c 1 + 2 ,

V˜n =

V˜t1 = Vt1 ,

V˜t2 = Vt2 , (55)

After some manipulations, ⎤ 1 ⎥ ⎢ ⎥ ⎢ −c˜ ↔  ↔  ⎥ ⎢ ρV n 1 ⎥ ⎢ ˜ ˜ Vt 1 = F · n + λ1 α˜ 1 K = F · n + (−c) ˜ ⎥ ⎢ c˜ ⎢ ⎥ Vt 2 ⎦ ⎣ 1 ρ (ρE + P ) ⎡

 ↔ ∗ F · n

Roe

⎡

⎤ 0 ⎢ ⎥ = ⎣ (ρVn2 + ρVn c˜ + P ) n ⎦ . 0

(56)

˜ 1 from [13]. This leads again to a with λ˜1 = V˜n − c˜ = −c, ˜ α1 = ρVn /c˜ and K ∗ definition of P # $ )  ∗ (γ − 1) 2 P Man , = 1 + γ Man Man + 1 + (57) P Roe 2 which fulfills the entropy inequality as long as  Man ≥ −

2 , 3−γ

for γ =

7 5

Man > −1.12 .

(58)

Thus, the Roe flux is entropy stable for shocks, but not for supersonic rarefactions.

EC-ES Fluxes We can also apply an entropy conservative (EC) flux that is used for interior element interfaces and add an entropy stable dissipation term (ES) to compute the boundary flux via the mirrored states (46). This is exactly the strategy proposed in Parsani

16

F. J. Hindenlang et al.

et al. [12] to get the boundary flux. Such an EC-ES flux is presented in Winters et al. [15]    ↔ ∗ ↔   1 F · n = FEC UL , UR · n − D H WR − WL ES 2

(59)

where D is a dissipation matrix and the matrix H w  u is carefully derived from the left and right states. Details are given in [15], where two approaches for the dissipation are distinguished. One is a Lax-Friedrichs-type dissipation, scaling with the maximum eigenvalue λmax = |Vn |+c (referred to as ‘EC-LF’). The other is a Roe-type dissipation computed via the eigenstructure of the matrix (D H) (referred to as ‘EC-Roe’). If we carefully insert the two mirrored boundary states into (59), we again get an equation for the modified pressure 

P∗ P

  = 1 + γ Man |Man | + 1

(60)

EC-LF

for the Lax-Friedrichs-type dissipation and 

P∗ P

= 1 + γ Man

(61)

EC-Roe

for the Roe-type dissipation. Both approaches lead to an entropy stable boundary flux when using a mirrored state. Note that the modified pressure of the EC-Roe flux (61) exactly matches the one of the HLL flux (53).

4 Discussion In the previous section we have shown conditions under which a specified wall flux is stable. In the linear analysis, the central numerical flux adds no dissipation and is neutrally stable. In the nonlinear analysis, entropy is not generated if the numerical wall pressure is equal to the internal pressure, P ∗ = P  . For upwinded approximations, the amount of energy or entropy dissipation depends on the normal Mach number. Since the boundary condition is only imposed weakly through the numerical flux, the normal Mach number will not be exactly zero except in the convergence limit. In fact, flow computations (especially steady state ones) are usually initiated with an impulsive start, where the initial state is a uniform flow, and the normal Mach number is not zero. This has proved over time to be very robust in practice. The analysis above gives an explanation why. In the linear analysis the dissipation due to imposing the boundary condition is proportional to the square of the normal Mach number. With an impulsive start initialization, this dissipation will be large. As the flow develops and the

Stability of Wall Boundary Condition Procedures

17

3

10

2

10

1

Δs

10

0

10

RP LF EC-LF HLL Roe ρc=1, γ =7/5

−1

10

−2

10

−3

Δs

10

−5

10

1

10

0

10

−1

10

−2

10

−3

−1

−4

−3

−2

−1

0

Man

1

2

3

4

5

RP LF EC-LF HLL Roe ρc=1, γ =7/5 −0.8

−0.6

−0.4

−0.2

0

0.2

0.4

0.6

0.8

1

Man Fig. 1 Entropy contribution s (62) produced by the wall boundary flux. RP refers to the exact Riemann problem (45), LF to (50), EC-LF to (60), HLL to (53) and Roe to (57). Plotted over the normal Mach number ranges |Man | ≤ 5 on the top and restricted to |Man | ≤ 1 on the bottom

boundary condition is better enforced, the dissipation reduces, going away only as the approximate solution converges. A similar effect is observed for the use of the different approximate Riemann solvers in the nonlinear analysis. In Fig. 1, we compare the entropy contribution 

s = (ρc)Man

P∗ −1 P

(62)

for the different wall boundary fluxes, over a range of normal Mach numbers for (ρc) = 1 and γ = 7/5. When the boundary condition is exactly fulfilled (Man = 0),

18

F. J. Hindenlang et al.

the entropy contribution is zero. For low normal Mach numbers, all fluxes have the same behavior. Compared to the exact Riemann problem (RP), the Lax-Friedrichs flux and the EC-LF flux always produce more entropy whereas the HLL flux produces less entropy for impinging velocities Man > 0. The results of HLLC and EC-Roe fluxes are not plotted, as they coincide with the HLL flux. As shown in the analysis, the Roe flux produces a negative entropy change for supersonic rarefactions, implying that it is not suitable for all flow configurations. Acknowledgements This work was supported by a grant from the Simons Foundation (#426393, David Kopriva). Gregor J. Gassner has been supported by the European Research Council (ERC) under the European Union’s Eights Framework Program Horizon 2020 with the research project Extreme, ERC grant agreement no. 714487. Florian Hindenlang thanks Eric Sonnendrücker and the Max-Planck Institute for Plasma Physics in Garching for their constant support.

References 1. Black, K.: A conservative spectral element method for the approximation of compressible fluid flow. Kybernetika 35(1), 133–146 (1999) 2. Chen, T., Shu, C.-W.: Entropy stable high order discontinuous Galerkin methods with suitable quadrature rules for hyperbolic conservation laws. J. Comput. Phys. 345, 427–461 (2017) 3. Fisher, T., Carpenter, M.H., Nordström, J., Yamaleev, N.K., Swanson, C.: Discretely conservative finite-difference formulations for nonlinear conservation laws in split form: theory and boundary conditions. J. Comput. Phys. 234, 353–375 (2013) 4. Gassner, G.J., Winters, A.R., Kopriva, D.A.: Split form nodal discontinuous Galerkin schemes with summation-by-parts property for the compressible Euler equations. J. Comput. Phys. 327, 39–66 (2016) 5. Gassner, G.J., Winters, A.R., Hindenlang, F.J., Kopriva, D.A.: The BR1 scheme is stable for the compressible Navier–Stokes equations. J. Sci. Comput. (2018) 6. Kopriva, D.A.: Implementing Spectral Methods for Partial Differential Equations. Scientific Computation. Springer, Berlin (2009) 7. Kopriva, D.A., Gassner, G.: On the quadrature and weak form choices in collocation type discontinuous Galerkin spectral element methods. J. Sci. Comput. 44(2), 136–155 (2010) 8. Kopriva, D.A., Gassner, G.J.: An energy stable discontinuous Galerkin spectral element discretization for variable coefficient advection problems. SIAM J. Sci. Comput. 36(4), A2076–A2099 (2014) 9. Nordström, J.: Conservative finite difference formulations, variable coefficients, energy estimates and artificial dissipation. J. Sci. Comput. 29(3), 375–404 (2006) 10. Nordström, J.: A roadmap to well posed and stable problems in computational physics. J. Sci. Comput. (2016). https://doi.org/10.1007/s10915-016-0303-9 11. Nordström, J., Svard, M.: Well-posed boundary conditions for the Navier–Stokes equations. SIAM J. Numer. Anal. 43(3), 1231–1255 (2005) 12. Parsani, M., Carpenter, M.H., Nielsen, E.J.: Entropy stable wall boundary conditions for the three-dimensional compressible Navier–Stokes equations. J. Comput. Phys. 292, 88–113 (2015) 13. Toro, E.F.: Riemann Solvers and Numerical Methods for Fluid Dynamics. Springer, Berlin (2009)

Stability of Wall Boundary Condition Procedures

19

14. van der Vegt, J.J.W., van der Ven, H.: Slip flow boundary conditions in discontinuous Galerkin discretizations of the Euler equations of gas dynamics. In Mang, H.A., Rammenstorfer, F.G. (eds.) Proceedings of the 5th World Congress on Computational Mechanics (WCCM V), number NLR-TP in Technical Publications, pp. 1–16. National Aerospace Laboratory, NLR (2002) 15. Winters, A.R., Derigs, D., Gassner, G.J., Walch, S.: A uniquely defined entropy stable matrix dissipation operator for high Mach number ideal MHD and compressible Euler simulations. J. Comput. Phys. 332, 274–289 (2017)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler Equations Florian J. Hindenlang and Gregor J. Gassner

1 Introduction Discontinuous Galerkin spectral element collocation method (DGSEM) with either Legendre-Gauss or Legendre-Gauss-Lobatto (LGL) nodes (see e.g. [14]) are among the most efficient variants in the class of element based high order methods, such as e.g. discontinuous Galerkin, flux reconstruction, or summation-by-parts (SBP) finite differences. In particular, the LGL variant, starting in [9], turned out to be similar to a SBP finite difference approximation with simultaneous-approximate-term technique (SAT). This relationship allowed to construct conservative skew-symmetric approximations, e.g. [9, 10, 21], and later enabled DGSEM-LGL approximations that are discretely entropy stable, e.g. [1, 3, 6, 8, 13, 17, 19, 20], and/or kinetic energy preserving [12]. These novel variants of nodal split form DG methods feature drastically increased non-linear robustness towards aliasing induced instabilities and favourable properties regarding the simulation of unresolved turbulence, e.g. [7, 23]. In addition to the very robust dissipative entropy stable versions, it is also possible to construct virtually dissipation free variants by choosing appropriate element interface numerical fluxes. These entropy conserving variants all show an odd-even behavior when experimentally testing the order of convergence, e.g. [9, 21], where the observed convergence order for even polynomial degrees

F. J. Hindenlang Max Planck Institute for Plasma Physics, Garching, Germany e-mail: [email protected] G. J. Gassner () Department for Mathematics and Computer Science, Center for Data and Simulation Science, University of Cologne, Cologne, Germany e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_2

21

22

F. J. Hindenlang and G. J. Gassner

N is N and for odd N is N + 1. Lately, a discussion emerged in the community, with interesting debates during the recent ICOSAHOM conference in London, where researchers reported non-optimal convergence behavior of the entropy stable DGSEM-LGL even with dissipative numerical surface fluxes, e.g. [6]. This paper contributes to this discussion and presents results of an experimental convergence order study for the compressible Euler equations with (1) the standard DGSEM with either Gauss and LGL nodes, (2) the entropy stable DGSEM with LGL nodes. For these nodal schemes, we test the convergence order with different numerical surface fluxes and report the results depending on the Mach number of the test case. The remainder of the paper is organized as follows: in the next section we describe the numerical model for our numerical experiments, in Sect. 3 we present our observed experimental convergence orders for different configurations and draw our conclusion in Sect. 4.

2 Numerical Model We consider the compressible Euler equations defined in the domain  ⊂ R3 ut +

3  ∂fi = 0. ∂xi

(1)

i=1

The state vector contains the conservative variables and the advective flux components are ⎤ ⎡ ⎡ ⎡ ⎤ ⎤ ⎤ v v v  1 2 3 ⎥ ⎢ ⎢ ⎢ ⎥ ⎥ ⎥ ⎢ ⎢ v 2 + p ⎥ ⎢ v v ⎥ ⎢ v v ⎥ ⎢ v ⎥  1⎥ 2 1 ⎥ 3 1 ⎥ ⎢ ⎢ ⎢ ⎥ 1 ⎢ →⎥ ⎢ ⎢ ⎢ ⎢ ⎥ ⎥ ⎥ ⎥ ⎢ v ⎥ u=⎢ ⎥ , f = ⎢ v1 v2 ⎥ , f2 = ⎢ v22 + p ⎥ , f3 = ⎢ v3 v2 ⎥ . ⎣ v ⎦ = ⎢ ⎢ ⎢ ⎥ ⎥ ⎥ ⎢ 2⎥ 1 ⎢ ⎢ v1 v3 ⎥ ⎢ v2 v3 ⎥ ⎢ v 2 + p ⎥ ⎢ v3 ⎥ E ⎦ ⎣ ⎣ ⎣ ⎦ ⎦ ⎦ ⎣ 3 E (E + p)v1 (E + p)v2 (E + p)v3 ⎡

⎡

⎤

(2) →

Here, , v = (v1 , v2 , v3 )T , p, E are the mass density, fluid velocities, pressure and total energy. We close the system with the ideal gas assumption, which relates the total energy and pressure 

1 → 2 p = (γ − 1) E −  v , 2 where γ denotes the adiabatic coefficient.

(3)

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . .

23

For our discretization, we subdivide the domain into non-overlapping hexahedral elements. For each element, we define a transfinite mapping to a unit reference space and use this mapping to transform the Eq. (1) from physical to reference space. A weak form is created by taking the inner product of the transformed equation with a test function. We use integration-by-parts for the flux term and approximate the resulting weak form as follows: the conservative variables are approximated by a polynomial in reference space with degree N, interpolated at the Gauss or LGL nodes. The volume fluxes are replaced by a standard interpolation of the non-linear flux function at the same Gauss/LGL nodes (standard DGSEM-Gauss or DGSEMLGL), see e.g. [14]. For the LGL variant, we are also able to introduce the split form volume integral based on entropy conserving and kinetic energy preserving numerical volume fluxes (Split-DGSEM), e.g. [12] and [22], resulting in either the entropy conserving or entropy stable DGSEM variants, depending on the choice of numerical surface flux.

3 Convergence Results In this section, we compare the convergence of the standard DGSEM and the entropy conservative and entropy stable discretization for different choices of the numerical flux and polynomial degrees N = 2, 3, 4, 5. We choose the test case of a two-dimensional density wave, with a constant pressure and transported with a constant velocity, which was proposed for onedimensional convergence tests in [4]. The density evolves as    (x1 , x2 , t) = 1 + 0.1 sin π (x1 − v1 t) + (x2 − v2 t)

(4)

with a prescribed velocity (v1 , v2 ). The pressure is chosen as p = 1/γ with γ = 1.4, so that the sound speed ranges between c = 0.95 . . . 1.05. Thus, by → changing the velocity, we change the Mach number of the flow Ma = |v|/c. Three Mach numbers are chosen: Ma ≈ 0.2 with (v1 , v2 ) = (0.1, 0.15), Ma ≈ 1.0 with (v1 , v2 ) = (0.7, 0.65) and Ma ≈ 3.5 with (v1 , v2 ) = (2.5, 2.4). The experimental order of convergence (EOC) is computed with the L2 error of the density at t = 1. The convergence study is performed with the open source, three-dimensional curvilinear split-form DG framework FLUXO (www.github.com/project-fluxo). As the test case is two-dimensional, we use fully periodic cartesian meshes of the domain [−1, 1]3 with an equal number of elements in x- and y-directions and always one element in z-direction. Note that h0 in the convergence tables refers to the coarsest mesh level, which is 42 elements for N = 2, 3 (h0 = 1/2) and 22 elements for N = 4, 5 (h0 = 1).

24

F. J. Hindenlang and G. J. Gassner

All simulation results are obtained with an explicit five stage, fourth order accurate low storage Runge–Kutta scheme [2], where a stable time step is computed according to the adjustable coefficient CF L ∈ (0, 1] the local maximum wave speed, and the relative grid size, e.g. [11]. We made sure that the time integrator did not influence the spatial convergence order, by adjusting the CFL number accordingly.

3.1 Standard DGSEM The convergence of the standard DGSEM with Gauss-Legendre nodes (DGSEMGauss) and with Legendre-Gauss-Lobatto (DGSEM-LGL) is shown in Tables 1 and 2, for the three Mach numbers and two choices of the numerical flux, namely the HLL (Harten, Lax, van Leer) flux and the Roe flux. The results of the LLF (local Lax-Friedrichs) flux and the HLLC flux (HLL variant with three waves, C for ‘contact’ wave) are reported in the Appendix, as the HLL results are similar to LLF, and HLLC behaves exactly the same as Roe, see Tables 4 and 5. Details on the properties and the implementation of the LLF, HLL, HLLC, and Roe fluxes are found in the book of Toro [18] and the references therein. For the HLL flux and the low Mach number Ma = 0.2, we observe an odd-even behavior with an order reduction for even polynomial degrees N = 2, 4. Also for Ma = 1.0, the convergence for even degrees is slightly affected, whereas for the high Mach number, all fluxes converge with full order. Comparing the L2 errors of the finest mesh for HLL and Roe for the low Mach number, HLL is less accurate for N = 2, 4 and more accurate for N = 3, 5. All numerical fluxes are approximate Riemann solvers, but the LLF and HLL only use the maximum wave speeds, whereas the HLLC and Roe also take the contact wave into account, and therefore keep the full order of the scheme for all Mach numbers for this test case.

3.2 Entropy Conservative and Entropy Stable DGSEM Now, we investigate the order reduction of the entropy conservative and entropy stable discretizations. Here, the standard DGSEM volume integral is replaced by split-form formulation (Split-DGSEM) using a two-point entropy conservative and kinetic energy preserving flux (ECKEP). If we choose the ECKEP flux at the surface, we get an entropy-conserving scheme. For entropy stability, we can use the LLF or HLL flux directly at the surface, or use the ECKEP flux and add a dissipation term, which must still satisfy the entropy inequality condition. In Winters et al. [22], such dissipation terms are carefully derived, using either only the maximum wave

DGSEM-Gauss + Roe

3.5

DGSEM-Gauss + HLL

1.0

3.5

0.2

1.0

Mach

Volume disc. + Surface flux

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

Mesh level

N =2 L2 () 1.87e−04 2.27e−05 2.82e−06 3.53e−07 2.60e−04 3.74e−05 4.88e−06 6.18e−07 4.87e−04 1.08e−04 1.95e−05 3.06e−06 1.87e−04 2.27e−05 2.82e−06 3.53e−07 1.82e−04 2.26e−05 2.82e−06 3.53e−07 EOC 3.34 3.04 3.00 3.00  2.55 2.80 2.94 2.98  1.70 2.18 2.46 2.67  3.34 3.04 3.00 3.00  3.07 3.00 3.00 3.00 

N =3 L2 () 8.57e−06 5.35e−07 3.34e−08 2.09e−09 5.92e−06 3.27e−07 1.95e−08 1.20e−09 4.36e−06 1.10e−07 5.84e−09 2.34e−10 8.57e−06 5.35e−07 3.34e−08 2.09e−09 8.76e−06 5.40e−07 3.35e−08 2.09e−09 EOC 4.02 4.00 4.00 4.00  4.55 4.18 4.07 4.03  5.05 5.31 4.23 4.64  4.02 4.00 4.00 4.00  4.04 4.02 4.01 4.00 

N =4 L2 () 1.03e−05 3.30e−07 1.02e−08 3.22e−10 1.15e−05 4.94e−07 1.73e−08 5.60e−10 1.57e−05 9.90e−07 5.06e−08 2.26e−09 1.03e−05 3.30e−07 1.02e−08 3.22e−10 1.11e−05 3.44e−07 1.05e−08 3.25e−10 EOC 5.02 4.96 5.01 4.99  4.39 4.54 4.84 4.95  2.56 3.99 4.29 4.49  5.02 4.96 5.01 4.99  5.07 5.00 5.04 5.01 

N =5 L2 () 6.76e−07 1.07e−08 1.66e−10 2.60e−12 6.74e−07 7.23e−09 9.98e−11 1.52e−12 9.80e−07 4.47e−09 3.96e−11 4.57e−13 6.76e−07 1.07e−08 1.66e−10 2.60e−12 6.95e−07 1.07e−08 1.67e−10 2.61e−12

Table 1 Experimental order of convergence of L2 error to the exact density (4), using the standard DGSEM-Gauss with HLL and Roe fluxes

(continued)

EOC 6.07 5.99 6.01 5.99  6.84 6.54 6.18 6.04  6.66 7.77 6.82 6.44  6.07 5.99 6.01 5.99  6.08 6.01 6.01 6.00 

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . . 25

Mesh level

h0 /2 h0 /4 h0 /8 h0 /16

Mach

0.2

N =2 L2 () 2.14e−04 2.22e−05 2.82e−06 3.53e−07

Full order is marked with  ( N + 1) and an order reduction with 

Volume disc. + Surface flux

Table 1 (continued) EOC 2.65 3.26 2.98 3.00 

N =3 L2 () 1.04e−05 5.49e−07 3.76e−08 2.07e−09 EOC 3.78 4.25 3.87 4.19 

N =4 L2 () 1.16e−05 3.49e−07 1.04e−08 3.38e−10 EOC 4.53 5.05 5.07 4.94 

N =5 L2 () 7.76e−07 1.08e−08 1.70e−10 2.64e−12

EOC 5.78 6.17 5.99 6.01 

26 F. J. Hindenlang and G. J. Gassner

DGSEM-LGL + Roe

3.5

DGSEM-LGL + HLL

1.0

3.5

0.2

1.0

Mach

Volume disc. + Surface flux

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

Mesh level

N =2 L2 () 1.22e−03 1.26e−04 1.48e−05 1.81e−06 1.04e−03 1.58e−04 2.11e−05 2.69e−06 1.20e−03 2.72e−04 5.57e−05 1.01e−05 1.22e−03 1.26e−04 1.48e−05 1.81e−06 9.17e−04 1.15e−04 1.44e−05 1.80e−06 EOC 3.38 3.27 3.10 3.03  2.30 2.72 2.91 2.97  1.96 2.14 2.29 2.47  3.38 3.27 3.10 3.03  2.86 2.99 3.00 3.00 

N =3 L2 () 3.85e−05 2.41e−06 1.51e−07 9.42e−09 3.44e−05 1.88e−06 1.16e−07 7.18e−09 4.86e−05 1.85e−06 1.21e−07 5.95e−09 3.85e−05 2.41e−06 1.51e−07 9.42e−09 3.96e−05 2.41e−06 1.51e−07 9.42e−09 EOC 4.03 4.00 4.00 4.00  4.44 4.20 4.02 4.01  3.51 4.71 3.94 4.34  4.03 4.00 4.00 4.00  3.94 4.04 4.00 4.00 

N =4 L2 () 4.34e−05 1.39e−06 4.34e−08 1.36e−09 4.11e−05 1.77e−06 6.25e−08 2.05e−09 5.16e−05 2.99e−06 1.56e−07 7.19e−09 4.34e−05 1.39e−06 4.34e−08 1.36e−09 4.41e−05 1.47e−06 4.38e−08 1.35e−09 EOC 4.93 4.96 5.01 4.99  4.89 4.54 4.82 4.93  3.91 4.11 4.26 4.44  4.93 4.96 5.01 4.99  4.94 4.90 5.07 5.02 

N =5 L2 () 2.70e−06 4.33e−08 6.67e−10 1.05e−11 2.89e−06 3.85e−08 5.37e−10 8.36e−12 3.85e−06 4.25e−08 5.45e−10 7.11e−12 2.70e−06 4.33e−08 6.67e−10 1.05e−11 2.76e−06 4.44e−08 6.84e−10 1.08e−11

Table 2 Experimental order of convergence of L2 error to the exact density (4), using standard DGSEM-LGL with HLL and Roe fluxes

(continued)

EOC 5.93 5.96 6.02 5.98  6.33 6.23 6.16 6.00  5.31 6.50 6.28 6.26  5.93 5.96 6.02 5.98  6.02 5.96 6.02 5.99 

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . . 27

Mesh level

h0 /2 h0 /4 h0 /8 h0 /16

Mach

0.2

N =2 L2 () 9.26e−04 1.19e−04 1.43e−05 1.80e−06

Full order is marked with  ( N + 1) and an order reduction with 

Volume disc. + Surface flux

Table 2 (continued) EOC 2.35 2.96 3.06 2.99 

N =3 L2 () 4.63e−05 2.40e−06 1.57e−07 9.37e−09 EOC 3.45 4.27 3.93 4.07 

N =4 L2 () 4.26e−05 1.59e−06 4.34e−08 1.49e−09 EOC 4.27 4.74 5.19 4.86 

N =5 L2 () 2.97e−06 4.48e−08 6.83e−10 1.09e−11

EOC 5.37 6.05 6.04 5.97 

28 F. J. Hindenlang and G. J. Gassner

Split-DGSEM + HLL

3.5

Split-DGSEM + ECKEP

1.0

3.5

0.2

1.0

Mach

Volume disc. + Surface flux

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

Mesh level

N =2 L2 () 1.62e−03 1.30e−04 1.05e−05 1.69e−06 1.41e−03 1.25e−04 1.48e−05 1.32e−06 1.13e−03 1.13e−04 1.12e−05 1.98e−06 1.23e−03 1.27e−04 1.48e−05 1.81e−06 1.04e−03 1.58e−04 2.11e−05 2.69e−06 EOC 4.23 3.64 3.62  2.64  3.89 3.49 3.09 3.48  2.26 3.32 3.34  2.50  3.36 3.27 3.10 3.03  2.30 2.72 2.90 2.97 

N =3 L2 () 8.45e−05 7.14e−06 7.90e−07 9.58e−08 9.45e−05 1.26e−05 1.60e−06 2.01e−07 8.03e−05 1.02e−05 1.29e−06 1.61e−07 3.88e−05 2.42e−06 1.51e−07 9.46e−09 3.48e−05 1.88e−06 1.16e−07 7.20e−09 EOC 2.77 3.56 3.18 3.04  2.34 2.90 2.98 3.00  2.98 2.97 2.99 3.00  4.06 4.00 4.00 4.00  4.44 4.21 4.02 4.01 

N =4 L2 () 5.80e−05 1.60e−06 4.56e−08 1.23e−09 7.71e−05 1.98e−06 4.06e−08 1.13e−09 5.95e−05 1.85e−06 4.02e−08 1.43e−09 4.49e−05 1.43e−06 4.44e−08 1.40e−09 4.16e−05 1.79e−06 6.39e−08 2.11e−09 EOC 5.88 5.18 5.13 5.21  5.86 5.28 5.61 5.17  4.14 5.01 5.52 4.81  4.94 4.97 5.01 4.99  4.90 4.54 4.81 4.92 

N =5 L2 () 4.63e−06 8.38e−08 3.59e−09 1.18e−10 3.49e−06 5.00e−08 1.22e−09 3.85e−11 4.11e−06 1.33e−07 4.21e−09 1.32e−10 3.01e−06 4.84e−08 7.48e−10 1.18e−11 3.48e−06 4.66e−08 6.08e−10 9.28e−12

(continued)

EOC 5.63 5.79 4.54 4.93  6.00 6.12 5.36 4.98  5.14 4.95 4.98 5.00  5.97 5.96 6.02 5.98  6.10 6.23 6.26 6.03 

Table 3 Experimental order of convergence of L2 error to the exact density (4), using entropy conservative ECKEP flux and entropy stable HLL and ECKEPRoe fluxes

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . . 29

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

0.2

0.2

1.0

3.5

Mesh level

Mach

N =2 L2 () 1.20e−03 2.72e−04 5.58e−05 1.01e−05 1.23e−03 1.27e−04 1.48e−05 1.81e−06 9.18e−04 1.15e−04 1.44e−05 1.80e−06 9.26e−04 1.19e−04 1.43e−05 1.80e−06

Full order is marked with  ( N + 1) and an order reduction with 

Split-DGSEM + ECKEP-Roe

Volume disc. + Surface flux

Table 3 (continued) EOC 1.96 2.14 2.29 2.47  3.36 3.27 3.10 3.03  2.86 2.99 3.00 3.00  2.35 2.96 3.06 2.99 

N =3 L2 () 4.90e−05 1.87e−06 1.21e−07 5.98e−09 3.88e−05 2.42e−06 1.51e−07 9.46e−09 3.98e−05 2.42e−06 1.51e−07 9.46e−09 4.65e−05 2.40e−06 1.58e−07 9.40e−09 EOC 3.51 4.71 3.95 4.34  4.06 4.00 4.00 4.00  3.94 4.04 4.00 4.00  3.45 4.27 3.93 4.07 

N =4 L2 () 5.23e−05 3.02e−06 1.57e−07 7.28e−09 4.49e−05 1.43e−06 4.44e−08 1.40e−09 4.51e−05 1.51e−06 4.49e−08 1.38e−09 4.37e−05 1.63e−06 4.45e−08 1.53e−09 EOC 3.92 4.12 4.26 4.43  4.94 4.97 5.01 4.99  4.92 4.90 5.07 5.02  4.26 4.74 5.19 4.86 

N =5 L2 () 4.47e−06 6.14e−08 6.18e−10 8.20e−12 3.01e−06 4.84e−08 7.48e−10 1.18e−11 3.08e−06 4.97e−08 7.66e−10 1.21e−11 3.26e−06 5.03e−08 7.66e−10 1.22e−11

EOC 5.16 6.19 6.64 6.23  5.97 5.96 6.02 5.98  5.95 5.95 6.02 5.99  5.34 6.02 6.04 5.97 

30 F. J. Hindenlang and G. J. Gassner

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . .

31

speed (LLF-type) or incorporating all waves (Roe-type), which we will refer to as ECKEP-LLF and ECKEP-Roe fluxes. In Table 3, we summarize the convergence of the dissipation-free ECKEP flux, the HLL and ECKEP-Roe flux. The results for LLF and ECKEP-LLF fluxes are found in the Appendix in Table 6, as they have the same convergence and error levels as the HLL flux. As expected, the dissipation-free surface flux (ECKEP) produces an order reduction for all Mach numbers for N = 3, 5, and for N = 2 full order is not kept in the last refinement step. If we simply use the HLL flux, we have an entropy stable scheme, but an order reduction for N = 2, 4 can be observed for the low Mach number flow, analogously to the standard DGSEM-LGL scheme. Interestingly, the odd-even behavior switches between entropy conserving and entropy stable fluxes. The ECKEP-Roe entropy stable flux accounts for all waves of the Riemann problem and adjusts the dissipation for each wave accordingly, which gives full order convergence for all Mach numbers.

4 Conclusions In this work, we report the convergence of standard DGSEM Gauss and GaussLobatto schemes to entropy conservative (EC) and entropy stable (ES) DGSEM schemes for the Euler equations, as there have been findings of order reduction for EC and ES schemes. We choose a simple density transport test case on a periodic domain and investigate the influence of the Mach number of the transport velocity. The EC scheme is dissipation free and an order reduction is observed by the convergence study presented here, confirming many similar observations found in literature. We also confirm that the ES scheme can have an order reduction for low Mach numbers, but only if the entropy stable numerical flux relies on simple approximate Riemann solvers such as local Lax-Friedrichs or HLL. If all waves are accounted for in the dissipation term of the entropy stable flux as presented in [22], the full order is observed for all Mach numbers. In addition, we reproduce the same behavior for the standard DGSEM Gauss and Gauss-Lobatto schemes, where the LLF and HLL fluxes suffer from order reduction at low Mach number, and HLLC and Roe fluxes have full order for all Mach numbers. We want to emphasize that the present convergence study should be seen merely as an observation, confirming that the numerical flux can have strong influence on the convergence order for both the standard DGSEM and the entropy stable DGSEM. Also, we stress that in our tests the order reduction is related to the form of the dissipation term in the numerical surface flux and is not related to the insufficient integration precision of the LGL-quadrature. Based on the observations presented in this work, a possible explanation for the loss of convergence for the density transport at low Mach numbers when using LLF and HLL fluxes is the form of dissipation from the approximate Riemann solver. In the case of the density transport, the exact solution follows the characteristic

32

F. J. Hindenlang and G. J. Gassner →

with velocity v. However, the approximate Riemann solver LLF and HLL consider → only two waves with maximum velocity ∼ (|v| + c) and do not consider the → contact wave with velocity v. Thus, the contact wave is dissipated proportional to → → → ∼ (|v| + c) and not to |v|. For low Mach numbers, where c > |v|, this causes over-upwinding. Over-upwinding was discussed in [5, 15]. It is not intuitive at first, but over-upwinding (over-penalization) can lead to a reduction of the in-built dissipation of the DG scheme, getting wave-propagation characteristics similar to a continuous Galerkin method [16]. This loss of in-built dissipation could be an explanation for the even-odd behavior we observed. However, it is still unclear why numerical surface fluxes with no in-built dissipation that are symmetric, e.g. EC flux, lead to an odd-even behavior in the convergence order and why numerical surface fluxes with over-upwinding, i.e. reduced dissipation due to over-penalization, cause an opposite even-odd behavior. What supports the explanation is the recovery of full convergence order for LLF and HLL when the difference in wave speed becomes smaller for higher Mach numbers, i.e. no over-upwinding. In contrast to LLF and HLL, the HLLC and Roe solvers take specifically the contact wave into account and adjust the dissipation accordingly and thus avoid strong over-upwinding by construction. In our tests, we always observe full convergence order for all Mach numbers for HLLC and Roe. Lastly we note that a convergence study using a manufactured solution technique can be misleading, as full convergence order is found independent of the choice of numerical flux. Hence, the introduction of a source term to balance the prescribed solution overcomes possible deficiencies of the surface fluxes, showing the limit of the manufactured solution technique in this context. In the Appendix, the convergence results of a manufactured solution are reported. Acknowledgements Gregor Gassner thanks the European Research Council for funding through the ERC Starting Grant “An Exascale aware and Un-crashable Space-Time-Adaptive Discontinuous Spectral Element Solver for Non-Linear Conservation Laws” (Extreme), ERC grant agreement no. 714487. Florian Hindenlang thanks Eric Sonnendrücker and the Max-Planck Institute for Plasma Physics in Garching for their constant support. We would also like to thank all participants of the ICOSAHOM 2018 for the valuable discussions on the topic of entropy stable schemes, which motivated this work.

Appendix Additional Convergence Results In this section, we present additional convergence results of the density wave test case for the DGSEM-Gauss and DGSEM-LGL with LLF and HLLC fluxes in Table 4 and Table 5, and also the entropy stable schemes with LLF and ECKEPLLF fluxes in Table 6. The results for LLF-type fluxes behave like the HLL flux, and for the HLLC flux like the Roe-type fluxes presented in Table 3.

DGSEM-Gauss + HLLC

3.5

DGSEM-Gauss + LLF

1.0

3.5

0.2

1.0

Mach

Volume disc. + Surface flux

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

Mesh level

N =2 L2 () 2.42e−04 3.24e−05 4.15e−06 5.22e−07 3.13e−04 5.30e−05 7.43e−06 9.61e−07 4.95e−04 1.12e−04 2.06e−05 3.29e−06 1.87e−04 2.27e−05 2.82e−06 3.53e−07 1.82e−04 2.26e−05 2.82e−06 3.53e−07 EOC 2.97 2.90 2.96 2.99  2.29 2.56 2.83 2.95  1.69 2.15 2.44 2.65  3.34 3.04 3.00 3.00  3.07 3.00 3.00 3.00 

N =3 L2 () 6.43e−06 3.71e−07 2.27e−08 1.41e−09 4.59e−06 2.25e−07 1.29e−08 7.65e−10 4.33e−06 1.06e−07 5.47e−09 2.15e−10 8.57e−06 5.35e−07 3.34e−08 2.09e−09 8.76e−06 5.40e−07 3.35e−08 2.09e−09 EOC 4.41 4.11 4.03 4.01  4.84 4.35 4.12 4.07  5.07 5.35 4.28 4.67  4.02 4.00 4.00 4.00  4.04 4.02 4.01 4.00 

N =4 L2 () 1.05e−05 4.32e−07 1.47e−08 4.73e−10 1.18e−05 6.08e−07 2.47e−08 8.53e−10 1.58e−05 1.01e−06 5.23e−08 2.38e−09 1.03e−05 3.30e−07 1.02e−08 3.22e−10 1.11e−05 3.44e−07 1.05e−08 3.25e−10 EOC 4.51 4.60 4.87 4.96  3.90 4.28 4.62 4.85  2.53 3.97 4.27 4.46  5.02 4.96 5.01 4.99  5.07 5.00 5.04 5.01 

Table 4 Experimental order of convergence of L2 error to the exact density (4), using DGSEM-Gauss with LLF and HLLC fluxes N =5 L2 () 6.68e−07 8.04e−09 1.15e−10 1.77e−12 6.69e−07 5.56e−09 6.79e−11 9.97e−13 9.88e−07 4.46e−09 3.86e−11 4.35e−13 6.76e−07 1.07e−08 1.66e−10 2.60e−12 6.95e−07 1.07e−08 1.67e−10 2.61e−12

(continued)

EOC 6.60 6.38 6.13 6.02  7.40 6.91 6.35 6.09  6.68 7.79 6.85 6.47  6.07 5.99 6.01 5.99  6.08 6.01 6.01 6.00 

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . . 33

Mesh level

h0 /2 h0 /4 h0 /8 h0 /16

Mach

0.2

N =2 L2 () 2.14e−04 2.22e−05 2.82e−06 3.53e−07

Full order is marked with  ( N + 1) and an order reduction with 

Volume disc. + Surface flux

Table 4 (continued) EOC 2.65 3.26 2.98 3.00 

N =3 L2 () 1.04e−05 5.49e−07 3.76e−08 2.07e−09 EOC 3.78 4.25 3.87 4.19 

N =4 L2 () 1.16e−05 3.49e−07 1.04e−08 3.38e−10 EOC 4.53 5.05 5.07 4.94 

N =5 L2 () 7.76e−07 1.08e−08 1.70e−10 2.64e−12

EOC 5.78 6.17 5.99 6.01 

34 F. J. Hindenlang and G. J. Gassner

DGSEM-LGL + HLLC

3.5

DGSEM-LGL + LLF

1.0

3.5

0.2

1.0

Mach

Volume disc. + Surface flux

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

Mesh level

N =2 L2 () 1.23e−03 1.51e−04 1.89e−05 2.36e−06 1.10e−03 2.03e−04 2.99e−05 3.94e−06 1.21e−03 2.77e−04 5.78e−05 1.06e−05 1.22e−03 1.26e−04 1.48e−05 1.81e−06 9.17e−04 1.15e−04 1.44e−05 1.80e−06 EOC 3.11 3.02 3.00 3.00  1.97 2.43 2.76 2.93  1.95 2.12 2.26 2.44  3.38 3.27 3.10 3.03  2.86 2.99 3.00 3.00 

N =3 L2 () 3.44e−05 2.00e−06 1.23e−07 7.66e−09 3.15e−05 1.68e−06 1.03e−07 6.39e−09 4.88e−05 1.86e−06 1.21e−07 5.98e−09 3.85e−05 2.41e−06 1.51e−07 9.42e−09 3.96e−05 2.41e−06 1.51e−07 9.42e−09 EOC 4.36 4.10 4.02 4.01  4.83 4.23 4.02 4.01  3.51 4.72 3.94 4.34  4.03 4.00 4.00 4.00  3.94 4.04 4.00 4.00 

N =4 L2 () 4.02e−05 1.61e−06 5.55e−08 1.79e−09 4.02e−05 1.98e−06 8.35e−08 2.96e−09 5.20e−05 3.05e−06 1.61e−07 7.57e−09 4.34e−05 1.39e−06 4.34e−08 1.36e−09 4.41e−05 1.47e−06 4.38e−08 1.35e−09 EOC 4.82 4.64 4.86 4.95  4.93 4.34 4.57 4.82  3.90 4.09 4.24 4.41  4.93 4.96 5.01 4.99  4.94 4.90 5.07 5.02 

Table 5 Experimental order of convergence of L2 error to the exact density (4), using DGSEM-GL with LLF and HLLC fluxes N =5 L2 () 2.91e−06 3.80e−08 5.46e−10 8.54e−12 2.85e−06 3.68e−08 4.86e−10 7.53e−12 3.89e−06 4.27e−08 5.47e−10 7.16e−12 2.70e−06 4.33e−08 6.67e−10 1.05e−11 2.76e−06 4.44e−08 6.84e−10 1.08e−11

(continued)

EOC 6.14 6.26 6.12 6.00  6.73 6.28 6.24 6.01  5.31 6.51 6.29 6.26  5.93 5.96 6.02 5.98  6.02 5.96 6.02 5.99 

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . . 35

Mesh level

h0 /2 h0 /4 h0 /8 h0 /16

Mach

0.2

N =2 L2 () 9.26e−04 1.19e−04 1.43e−05 1.80e−06

Full order is marked with  ( N + 1) and an order reduction with 

Volume disc. + Surface flux

Table 5 (continued) EOC 2.35 2.96 3.06 2.99 

N =3 L2 () 4.63e−05 2.40e−06 1.57e−07 9.37e−09 EOC 3.45 4.27 3.93 4.07 

N =4 L2 () 4.26e−05 1.59e−06 4.34e−08 1.49e−09 EOC 4.27 4.74 5.19 4.86 

N =5 L2 () 2.97e−06 4.48e−08 6.83e−10 1.09e−11

EOC 5.37 6.05 6.04 5.97 

36 F. J. Hindenlang and G. J. Gassner

Split-DGSEM + ECKEP-LLF

3.5

Split-DGSEM + LLF

1.0

3.5

0.2

1.0

Mach

Volume disc. + Surface flux

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

Mesh level

N =2 L2 () 1.24e−03 1.52e−04 1.89e−05 2.36e−06 1.09e−03 2.03e−04 2.99e−05 3.94e−06 1.21e−03 2.77e−04 5.79e−05 1.06e−05 1.24e−03 1.52e−04 1.89e−05 2.36e−06 1.10e−03 2.04e−04 3.00e−05 3.95e−06 EOC 3.11 3.03 3.00 3.00  1.97 2.43 2.76 2.93  1.95 2.12 2.26 2.44  3.10 3.03 3.01 3.00  1.97 2.43 2.76 2.93 

N =3 L2 () 3.49e−05 2.01e−06 1.23e−07 7.68e−09 3.21e−05 1.69e−06 1.04e−07 6.41e−09 4.92e−05 1.88e−06 1.22e−07 6.01e−09 3.49e−05 2.01e−06 1.23e−07 7.68e−09 3.20e−05 1.69e−06 1.04e−07 6.41e−09 EOC 4.37 4.12 4.03 4.01  4.82 4.25 4.03 4.02  3.50 4.71 3.95 4.34  4.37 4.12 4.03 4.01  4.83 4.25 4.03 4.02 

N =4 L2 () 4.16e−05 1.64e−06 5.67e−08 1.84e−09 4.07e−05 1.99e−06 8.46e−08 3.03e−09 5.26e−05 3.07e−06 1.63e−07 7.66e−09 4.16e−05 1.64e−06 5.68e−08 1.84e−09 4.08e−05 2.00e−06 8.49e−08 3.04e−09 EOC 4.85 4.67 4.85 4.95  4.94 4.35 4.56 4.80  3.91 4.10 4.24 4.41  4.84 4.67 4.85 4.95  4.94 4.35 4.56 4.81 

Table 6 Experimental order of convergence of L2 error to the exact density (4), using entropy stable LLF and ECKEP-LLF flux N =5 L2 () 3.37e−06 4.49e−08 6.21e−10 9.58e−12 3.77e−06 4.69e−08 5.57e−10 8.38e−12 4.52e−06 6.21e−08 6.21e−10 8.26e−12 3.38e−06 4.49e−08 6.21e−10 9.58e−12 3.77e−06 4.69e−08 5.57e−10 8.38e−12

(continued)

EOC 6.05 6.23 6.18 6.02  6.34 6.33 6.39 6.05  5.15 6.19 6.64 6.23  6.06 6.23 6.18 6.02  6.36 6.33 6.40 6.05 

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . . 37

Mesh level

h0 /2 h0 /4 h0 /8 h0 /16

Mach

0.2

N =2 L2 () 1.21e−03 2.78e−04 5.79e−05 1.07e−05

Full order is marked with  ( N + 1) and an order reduction with 

Volume disc. + Surface flux

Table 6 (continued) EOC 1.95 2.12 2.26 2.44 

N =3 L2 () 4.93e−05 1.88e−06 1.22e−07 6.01e−09 EOC 3.51 4.71 3.95 4.34 

N =4 L2 () 5.27e−05 3.08e−06 1.63e−07 7.67e−09 EOC 3.91 4.10 4.24 4.41 

N =5 L2 () 4.53e−06 6.21e−08 6.21e−10 8.26e−12

EOC 5.15 6.19 6.65 6.23 

38 F. J. Hindenlang and G. J. Gassner

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . .

39

Manufactured Solution with Source Term Here, we run a convergence test with the method of manufactured solutions. To do so, we assume a two-dimensional solution of the form "T !  T u =  , v1 , v2 , v3 , E = g , g , g , 0 , g 2

(5)

with g = g(x1 , x2 , t) = 0.5 sin(2π(x1 + x2 − t)) + 2. Note that the average Mach number in the domain is Ma = 0.8. Inserting (5) into the Euler equations, and using the fact that spatial and time derivatives are g  = ∂x1 g = ∂x2 g = −∂t g, we get an additional residual ⎞ g ⎜ (3γ − 2)g  + 2(γ − 1)gg  ⎟ 3 ⎟ ⎜  ∂fi ⎟ ⎜ ut + = ⎜ (3γ − 2)g  + 2(γ − 1)gg  ⎟ ⎟ ⎜ ∂xi i=1 ⎠ ⎝ 0   (6γ − 2)g + 2(2γ − 1)gg ⎛

(6)

To solve the inhomogeneous problem, we subtract the residual from the approximate solution in each Runge–Kutta step. Moreover, we run the test case up to the final time t = 1.0. In the convergence results for the standard DGSEM Gauss and Gauss-Lobatto, we see that the LLF flux still leads to an order reduction for N = 2, 4, whereas full order is found for the HLL, HLLC and Roe fluxes, see Tables 7 and 8. In Table 9 the entropy conservative scheme shows again an order reduction for N = 3, 5, and the LLF-Type dissipation too, for N = 2, 4, and for this test case, all entropy stable schemes exhibit full order.

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

DGSEM-Gauss + LLF

N =2 L2 () 2.30e−03 4.81e−04 9.48e−05 1.57e−05 1.24e−03 1.17e−04 1.41e−05 1.76e−06 1.24e−03 1.17e−04 1.41e−05 1.76e−06 1.24e−03 1.17e−04 1.41e−05 1.76e−06 EOC 2.20 2.25 2.34 2.60  2.84 3.41 3.04 3.00  2.84 3.41 3.04 3.00  2.84 3.41 3.04 3.00 

Full order is marked with  ( N + 1) and an order reduction with 

DGSEM-Gauss + Roe

DGSEM-Gauss + HLLC

DGSEM-Gauss + HLL

Mesh level

Volume disc. + Surface flux

N =3 L2 () 4.54e−05 1.99e−06 1.02e−07 6.25e−09 5.46e−05 3.36e−06 1.85e−07 1.07e−08 5.46e−05 3.36e−06 1.85e−07 1.07e−08 5.46e−05 3.36e−06 1.85e−07 1.07e−08 EOC 5.34 4.52 4.28 4.03  4.35 4.02 4.18 4.11  4.35 4.02 4.18 4.11  4.35 4.02 4.18 4.11 

N =4 L2 () 1.13e−04 4.78e−06 2.88e−07 1.73e−08 1.32e−04 2.89e−06 7.14e−08 2.15e−09 1.32e−04 2.89e−06 7.14e−08 2.15e−09 1.32e−04 2.89e−06 7.14e−08 2.15e−09 EOC 6.35 4.56 4.05 4.05  5.22 5.51 5.34 5.05  5.22 5.51 5.34 5.05  5.22 5.51 5.34 5.05 

N =5 L2 () 4.52e−05 2.37e−07 1.53e−09 1.57e−11 1.47e−05 1.44e−07 1.74e−09 2.24e−11 1.47e−05 1.44e−07 1.74e−09 2.24e−11 1.47e−05 1.44e−07 1.74e−09 2.24e−11

EOC 4.38 7.58 7.28 6.61  6.30 6.67 6.37 6.28  6.30 6.67 6.37 6.28  6.30 6.67 6.37 6.28 

Table 7 Experimental order of convergence of L2 error of density for the manufactured solution (5), using DGSEM-Gauss with LLF, HLL, HLLC and Roe fluxes

40 F. J. Hindenlang and G. J. Gassner

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

DGSEM-LGL + LLF

N =2 L2 () 7.36e−03 1.33e−03 2.79e−04 5.31e−05 5.32e−03 5.99e−04 7.25e−05 9.02e−06 5.32e−03 5.99e−04 7.25e−05 9.02e−06 5.32e−03 5.99e−04 7.25e−05 9.02e−06 EOC 2.85 2.47 2.25 2.39  3.16 3.15 3.05 3.01  3.16 3.15 3.05 3.01  3.16 3.15 3.05 3.01 

Full order is marked with  ( N + 1) and an order reduction with 

DGSEM-LGL + Roe

DGSEM-LGL + HLLC

DGSEM-LGL + HLL

Mesh level

Volume disc. + Surface flux

N =3 L2 () 3.15e−04 1.43e−05 7.99e−07 4.72e−08 2.52e−04 1.38e−05 7.69e−07 4.74e−08 2.52e−04 1.38e−05 7.69e−07 4.74e−08 2.52e−04 1.38e−05 7.69e−07 4.74e−08 EOC 4.38 4.46 4.16 4.08  4.00 4.19 4.17 4.02  4.00 4.19 4.17 4.02  4.00 4.19 4.17 4.02 

N =4 L2 () 5.69e−04 2.04e−05 8.81e−07 5.94e−08 3.84e−04 1.43e−05 2.92e−07 7.77e−09 3.84e−04 1.43e−05 2.92e−07 7.77e−09 3.84e−04 1.43e−05 2.92e−07 7.77e−09 EOC 5.78 4.80 4.54 3.89  5.36 4.75 5.61 5.23  5.36 4.75 5.61 5.23  5.36 4.75 5.61 5.23 

N =5 L2 () 9.33e−05 9.27e−07 9.34e−09 1.37e−10 4.33e−05 4.58e−07 7.08e−09 1.10e−10 4.33e−05 4.58e−07 7.08e−09 1.10e−10 4.33e−05 4.58e−07 7.08e−09 1.10e−10

EOC 5.34 6.65 6.63 6.09  6.24 6.56 6.02 6.01  6.24 6.56 6.02 6.01  6.24 6.56 6.02 6.01 

Table 8 Experimental order of convergence of L2 error of density for the manufactured solution (5), using DGSEM-LGL with LLF, HLL, HLLC and Roe fluxes

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . . 41

h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16 h0 /2 h0 /4 h0 /8 h0 /16

Split-DGSEM + ECKEP

N =2 L2 () 1.31e−02 1.30e−03 1.23e−04 1.76e−05 7.60e−03 1.63e−03 3.40e−04 6.14e−05 7.66e−03 1.64e−03 3.41e−04 6.16e−05 5.82e−03 7.06e−04 8.63e−05 1.08e−05 5.81e−03 7.06e−04 8.63e−05 1.08e−05 EOC 3.53 3.34 3.40 2.80  3.13 2.22 2.26 2.47  3.21 2.23 2.26 2.47  3.10 3.04 3.03 3.00  3.11 3.04 3.03 3.00 

Full order is marked with  ( N + 1) and an order reduction with 

Split-DGSEM + ECKEP-Roe

Split-DGSEM + HLL

Split-DGSEM + ECKEP-LLF

Split-DGSEM + LLF

Mesh level

Volume disc. + Surface flux

N =3 L2 () 1.28e−03 1.13e−04 1.24e−05 1.67e−06 3.70e−04 1.90e−05 9.89e−07 6.41e−08 3.74e−04 1.90e−05 9.88e−07 6.40e−08 3.01e−04 2.04e−05 1.16e−06 7.20e−08 3.01e−04 2.04e−05 1.16e−06 7.20e−08 EOC 2.88 3.49 3.20 2.89  3.91 4.29 4.26 3.95  3.98 4.30 4.26 3.95  3.85 3.88 4.14 4.01  3.85 3.88 4.14 4.01 

N =4 L2 () 5.62e−03 3.12e−04 2.00e−06 3.41e−08 6.56e−04 3.09e−05 1.83e−06 9.15e−08 6.79e−04 3.08e−05 1.83e−06 9.17e−08 5.11e−04 1.67e−05 5.08e−07 1.62e−08 5.11e−04 1.67e−05 5.08e−07 1.62e−08 EOC 2.49 4.17 7.29 5.87  4.44 4.41 4.08 4.32  4.45 4.46 4.07 4.32  4.35 4.94 5.04 4.97  4.35 4.94 5.04 4.97 

N =5 L2 () 1.12e−03 5.97e−06 9.45e−08 3.17e−09 1.28e−04 2.01e−06 1.80e−08 2.17e−10 1.33e−04 2.06e−06 1.79e−08 2.17e−10 7.06e−05 1.08e−06 1.67e−08 2.64e−10 7.06e−05 1.08e−06 1.67e−08 2.64e−10

EOC 4.05 7.56 5.98 4.90  4.75 6.00 6.80 6.37  4.75 6.01 6.85 6.37  5.27 6.03 6.02 5.98  5.27 6.03 6.02 5.98 

Table 9 Experimental order of convergence of L2 error of density for the manufactured solution (5), using entropy conservative and entropy stable schemes

42 F. J. Hindenlang and G. J. Gassner

On the Order Reduction of Entropy Stable DGSEM for the Compressible Euler. . .

43

References 1. Bohm, M., Winters, A.R., Gassner, G.J., Derigs, D., Hindenlang, F., Saur, J.: An entropy stable nodal discontinuous Galerkin method for the resistive MHD equations. Part I: Theory and numerical verification. J. Comput. Phys. 108076 (2018). https://doi.org/10.1016/j.jcp.2018.06. 027 2. Carpenter, M., Kennedy, C.: Fourth-order 2N-storage Runge-Kutta schemes. Technical Report NASA TM 109111 (1994) 3. Carpenter, M., Fisher, T., Nielsen, E., Frankel, S.: Entropy stable spectral collocation schemes for the Navier–Stokes equations: discontinuous interfaces. SIAM J. Sci. Comput. 36(5), B835– B867 (2014) 4. Chan, J.: On discretely entropy conservative and entropy stable discontinuous Galerkin methods. J. Comput. Phys. 362, 346–374 (2018) 5. Chan, J., Warburton, T.: On the penalty stabilization mechanism for upwind discontinuous Galerkin formulations of first order hyperbolic systems. Comput. Math. Appl. 74(12), 3099– 3110 (2017) 6. Chen, T., Shu, C.-W.: Entropy stable high order discontinuous Galerkin methods with suitable quadrature rules for hyperbolic conservation laws. J. Comput. Phys. 345, 427–461 (2017) 7. Flad, D., Gassner, G.: On the use of kinetic energy preserving DG-schemes for large eddy simulation. J. Comput. Phys. 350, 782–795 (2017) 8. Friedrich, L., Winters, A.R., Del Rey Fernández, D.C., Gassner, G.J., Parsani, M., Carpenter, M.H.: An entropy stable h/p non-conforming discontinuous Galerkin method with the summation-by-parts property. J. Sci. Comput. 77, 689–725 (2018) 9. Gassner, G.: A skew-symmetric discontinuous Galerkin spectral element discretization and its relation to SBP-SAT finite difference methods. SIAM J. Sci. Comput. 35(3), A1233–A1253 (2013) 10. Gassner, G.J.: A kinetic energy preserving nodal discontinuous Galerkin spectral element method. Int. J. Numer. Methods Fluids 76(1), 28–50 (2014) 11. Gassner, G., Kopriva, D.A.: A comparison of the dispersion and dissipation errors of Gauss and Gauss–Lobatto discontinuous Galerkin spectral element methods. SIAM J. Sci. Comput. 33, 2560–2579 (2011) 12. Gassner, G.J., Winters, A.R., Kopriva, D.A.: Split form nodal discontinuous Galerkin schemes with summation-by-parts property for the compressible Euler equations. J. Comput. Phys. 327, 39–66 (2016) 13. Gassner, G.J., Winters, A.R., Hindenlang, F.J., Kopriva, D.A.: The BR1 scheme is stable for the compressible Navier–Stokes equations. J. Sci. Comput. 77, 154–200 (2018) 14. Kopriva, D.A.: Implementing Spectral Methods for Partial Differential Equations. Scientific Computation. Springer, Berlin (2009) 15. Manzanero, J., Ferrer, E., Rubio, G., Valero, E.: On the role of numerical dissipation in stabilising under-resolved turbulent simulations using discontinuous Galerkin methods. Preprint arXiv:1805.10519 (2018) 16. Moura, R., Sherwin, S., Peirò, J.: Eigensolution analysis of spectral/hp continuous Galerkin approximations to advection-diffusion problems: insights into spectral vanishing viscosity. J. Comput. Phys. 307, 401–422 (2016) 17. Parsani, M., Carpenter, M.H., Nielsen, E.J.: Entropy stable wall boundary conditions for the three-dimensional compressible Navier–Stokes equations. J. Comput. Phys. 292, 88–113 (2015) 18. Toro, E.: Riemann Solvers and Numerical Methods for Fluid Dynamics. Springer, Berlin (2009) 19. Wintermeyer, N., Winters, A.R., Gassner, G.J., Kopriva, D.A.: An entropy stable nodal discontinuous Galerkin method for the two dimensional shallow water equations on unstructured curvilinear meshes with discontinuous bathymetry. J. Comput. Phys. (2016, submitted). arXiv:1509.07096 [math.NA]

44

F. J. Hindenlang and G. J. Gassner

20. Wintermeyer, N., Winters, A.R., Gassner, G.J., Warburton, T.: An entropy stable discontinuous Galerkin method for the shallow water equations on curvilinear meshes with wet/dry fronts accelerated by GPUs. J. Comput. Phys. 375, 447–480 (2018) 21. Winters, A.R., Gassner, G.J.: A comparison of two entropy stable discontinuous Galerkin spectral element approximations for the shallow water equations with non-constant topography. J. Comput. Phys. 301, 357–376 (2015) 22. Winters, A.R., Derigs, D., Gassner, G.J., Walch, S.: A uniquely defined entropy stable matrix dissipation operator for high Mach number ideal MHD and compressible Euler simulations. J. Comput. Phys. 332, 274–289 (2017) 23. Winters, A.R., Moura, R.C., Mengaldo, G., Gassner, G.J., Walch, S., Peiro, J., Sherwin, S.J.: A comparative study on polynomial dealiasing and split form discontinuous Galerkin schemes for under-resolved turbulence computations. J. Comput. Phys. 372, 1–21 (2018)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

A Review of Regular Decompositions of Vector Fields: Continuous, Discrete, and Structure-Preserving Ralf Hiptmair and Clemens Pechstein

1 Introduction For a bounded Lipschitz domain Ω ⊂ R3 recall the classical L2 -orthogonal Helmholtz decompositions L2 (Ω) = ∇ H01 (Ω) ⊕ H(div 0, Ω) = ∇ H 1 (Ω) ⊕ H0 (div 0, Ω) , see, e.g., [9, Ch. XI, Sect. I]. They can be used to derive decompositions of (subspaces of) H(curl, Ω): H0 (curl, Ω) = ∇ H01 (Ω) ⊕ XN (Ω), XN (Ω) := H0 (curl, Ω) ∩ H(div 0, Ω), H(curl, Ω) = ∇ H 1 (Ω) ⊕ XT (Ω),

XT (Ω) := H(curl, Ω) ∩ H0 (div 0, Ω) .

If the domain Ω is convex then the respective complementary space, XN (Ω) or XT (Ω), is continuously embedded in the space H1 (Ω) of vector fields with Cartesian components in H 1 (Ω), cf. [1]. Then one can, for instance, write any u ∈ H(curl, Ω) as u = ∇ p + z,

(1)

R. Hiptmair () Seminar for Applied Mathematics, ETH Zürich, Zürich, Switzerland e-mail: [email protected] C. Pechstein Dassault Systèmes, Darmstadt, Germany e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_3

45

46

R. Hiptmair and C. Pechstein

with p ∈ H 1 (Ω) and z ∈ H1 (Ω). Since  ∇ pL2 (Ω) ≤ uL2 (Ω) one obtains (using the continuous embedding) the stability property1  ∇ pL2 (Ω) + zH1 (Ω) ≤ CzH(curl,Ω) .

(2)

A similar decomposition can be found for u ∈ H0 (curl, Ω). Generally, a decomposition of form (1) with the stability property (2) is called regular decomposition, even if L2 -orthogonality does not hold. Actually, it turns out that (1)–(2) can be achieved even in cases where Ω is non-convex, in particular on non-smooth domains, or in cases where Ω or its boundary have non-trivial topology; only the L2 -orthogonality has to be sacrificed, cf. [20]. Noting that ∇ H 1 (Ω) is contained in the kernel of the curl operator and that— under mild smoothness assumptions on the domain—the whole kernel is spanned by ∇ H 1 (Ω) plus a finite-dimensional co-homology space [15, Sect. 4] one can achieve a second decomposition, u = h +z,

(3)

with h ∈ ker(curl|H(curl,Ω) ) and z ∈ H1 (Ω), where hL2 (Ω) ≤ C uL2 (Ω) ,

zH1 (Ω) ≤ C  curl uL2 (Ω) .

(4)

The second stability estimate states that if u is already in the kernel of the curl operator, then z is zero. Hence, (1) the operator mapping u to h is a projection onto the kernel space and (2) the complement operator projects u to the function z of higher regularity H1 (Ω). For trivial topology of Ω and ∂Ω, the two decompositions (1)– (2) and (3)–(4) coincide. As a few among many more [17, Sect. 1.5], we would like to highlight two important applications of these regular decompositions. 1. The second form (3)–(4), in the sequel called rotation-bounded decomposition, can be used to show that the operator underlying a certain boundary value problem for Maxwell’s equations is a Fredholm operator. The key point is that the complement space of the kernel (from the view of the mentioned projections) is H1 (Ω) which is compactly embedded in L2 (Ω), see e.g., [14, 16] and references therein. 2. The first form (1)–(2), in the sequel called gradient-based decomposition, has been used to generate stable three-term splittings of a finite element subspace of H(curl, Ω), cf. [19–21, 23], which allows the construction of so-called fictitious or auxiliary space preconditioners for the ill-conditioned system matrix underlying the discretized Maxwell equations.

1 Here and below C stands for a positive “generic constant” that may depend only on Ω, unless specified otherwise.

Regular Decompositions

47

In both applications, it is desirable to obtain the decompositions for minimal smoothness of the domain, e.g., Lipschitz domains, which are not necessarily convex. Moreover, it is also desirable to go beyond decompositions of the entire space H(curl, Ω) and extend them to subspaces for which the appropriate trace vanishes on a “Dirichlet part” ΓD of the boundary. In this case traces of the two summands should also vanish on ΓD . In the present paper, we provide regular decompositions of both types for subspaces of H(curl, Ω) (in Sect. 3) and H(div, Ω) (in Sect. 4) comprising functions with vanishing trace on a part ΓD of the boundary ∂Ω for Lipschitz domains Ω of arbitrary topology. In particular, Ω is allowed to have handles, and ∂Ω and ΓD may have several connected components. The Dirichlet boundary ΓD must satisfy a certain smoothness assumption that we shall introduce in Sect. 2. In addition to the stability estimates (2) and (4), we show that the decompositions are stable even in L2 (Ω). In the final part of the manuscript, in Sect. 5, we establish regular decompositions of spaces of Whitney forms, which are lowest-order conforming finite element subspaces of H(curl, Ω) and H(div, Ω), respectively, built upon simplicial triangulations of Ω. This note is based on [17] and is an abridged version of [18]. Please refer to this latter preprint for complete proofs of the results quoted below.

2 Preliminaries Since subtle geometric arguments will play a major role for parts of the theory, we start with a precise characterization of the geometric setting: Let Ω ⊂ R3 be an open, bounded, connected Lipschitz domain.2 We write d(Ω) for its diameter. Its boundary Γ := ∂Ω is partitioned according to Γ = ΓD ∪ Σ ∪ ΓN , with relatively open sets ΓD and ΓN . We assume that this provides a piecewise C 1 dissection of ∂Ω in the sense of [12, Definition 2.2]. Sloppily speaking, this means that Σ is the union of closed curves that are piecewise C 1 . Under the above assumptions on Ω and ΓD , [12, Lemma 4.4] guarantees the existence of an open Lipschitz neighborhood ΩΓ (“Lipschitz collar”) of Γ and of a “bulge” ΥD ⊂ ΩΓ \ Ω. We recall the properties of bulge domains from [12, Sect. 2, Thm. 2.3], also stated in [17, Thm. 2.2]: Theorem 1 (Bulge-Augmented Domain) There exists a Lipschitz domain ΥD ⊂ R3 \Ω, such that Υ D ∩Ω = ΓD , Ω e := ΥD ∪ΓD ∪Ω is Lipschitz, d(Ω e ) ≤ 2 d(Ω), and Υ D ⊂ ΩΓ . Moreover, each connected component ΓD,k of ΓD corresponds to a connected component ΥD,k of ΥD , and these have positive distance from each other.

2 Strongly Lipschitz, in the sense that the boundary is locally the graph of a Lipschitz continuous function.

48

R. Hiptmair and C. Pechstein

Let HΓ1D (Ω) := {u ∈ H 1 (Ω) : (γ u)|ΓD = 0}, HΓD (curl, Ω) := {u ∈ H(curl, Ω) : (γ τ u)|ΓD = 0}, HΓD (div, Ω) := {u ∈ H(div, Ω) : (γn u)|ΓD = 0} , denote the standard Sobolev spaces where the distributional gradient, curl, or divergence is in L2 and where the pointwise trace γ u, the tangential trace γτ u, or the normal trace γn u, respectively, vanishes on the Dirichlet boundary ΓD , see e.g. [3, 6, 26]. These space are linked via the de Rham complex, ∇

id

curl

div

KΓD (Ω) −→ HΓ1D (Ω) −→ HΓD (curl, Ω) −→ HΓD (div, Ω) −→ L2 (Ω), (5) where ⎧ ⎨span{1}, if Γ = ∅, D KΓD (Ω) := {v ∈ HΓ1D (Ω) : v = const} = ⎩{0}, otherwise. The range of each operator in (5) lies in the kernel space of the succeeding one, cf. [3, Lemma 2.2]. We define HΓD (curl 0, Ω) := {v ∈ HΓD (curl, Ω) : curl v = 0}, HΓD (div 0, Ω) := {v ∈ HΓD (div, Ω) : div v = 0}.

(6)

Barring topological obstructions these kernels can be represented through potentials: Let β1 (Ω) denote the first Betti number of Ω (the number of “handles”) and β2 (Ω) the second Betti number (the number of connected components of ∂Ω minus one). By the very definition of the Betti numbers as dimensions of co-homology spaces we have β1 (Ω) = 0

⇒

β2 (Ω) = 0

⇒

H(curl 0, Ω) = ∇ H 1 (Ω), H(div 0, Ω) = curl H(curl, Ω),

(7) (8)

cf. [26]. We call Ω topologically trivial if β1 (Ω) = β2 (Ω) = 0.

3 Regular Decompositions and Potentials Related to H(curl) Throughout we rely on the properties of Ω and ΓD as introduced in Sect. 2 and use the notations from Theorem 1. We write C for positive “generic constants” and

Regular Decompositions

49

say that a constant “depends only on the shape of Ω and ΓD ”, if it depends on the geometric setting alone, but is invariant with respect to similarity transformations. To achieve this the diameter of Ω will have to enter the estimates; we denote it by d(Ω).

3.1 Gradient-Based Regular Decomposition of H(curl) The following theorem is essentially [17, Thm. 2.1]. Theorem 2 (Gradient-Based Regular Decomposition of H(curl)) Let (Ω, ΓD ) satisfy the assumptions of Sect. 2. Then for each u ∈ HΓD (curl, Ω) there exist z ∈ H1ΓD (Ω) and p ∈ HΓ1D (Ω) depending linearly on u such that u = z + ∇ p,

(i) (ii) (iii)

z0,Ω +  ∇ p0,Ω ≤ C u0,Ω ,  ∇ z0,Ω +

1 1 z0,Ω ≤ C curl u0,Ω + u0,Ω , d(Ω) d(Ω)

with constants depending only on the shape of Ω and ΓD , but not on d(Ω). Remark 1 An early decomposition of a subspace of H(curl, Ω) ∩ H(div, Ω) into a regular part in H1 (Ω) and a singular part in ∇H 1 (Ω) can be found in [4] and in [5, Proposition 5.1], see also [7, Sect. 3] and references therein. Theorem 2 was proved in [14, Lemma 2.4] for the case of ΓD = ∂Ω and without the L2 stability estimate, following [5, Proposition 5.1]. Pasciak and Zhao [28, Lemma 2.2] provided a version for simply connected Ω and the case ΓD = ∂Ω with pure L2 -stability, but p is only constant on each connected component of ∂Ω (see also Theorem 5 and Remark 3). This result was refined in [24, Thm. 3.1]. For the case ΓD = ∅, [14, Lemma 2.4] gives a similar decomposition but ∇p must be replaced by an element from H(curl 0, Ω) in general. Finally, Theorem 2 without the pure L2 -stability was proved in [20, Thm. 5.2].3 Remark 2 The constant C in Theorem 2 depends mainly on the stability constants of key extension operators. If the bulge ΥD has multiple components ΥD,k , the final estimate will depend on the relative distances between ΥD,k , ΥD, , k =  and the ratios d(ΥD,k )/ d(Ω).

3 This reference contains a typo which is easily identified when inspecting the proof: In general, z cannot be estimated in terms of  curl u0,Ω but one must use the full H(curl) norm.

50

R. Hiptmair and C. Pechstein

Remark 3 If ΓD = ∂Ω, one obtains only p ∈ H 1 (Ω) being constant on each connected component of ΓD but the improved bound  ∇ z0,Ω + d(Ω)−1 z0,Ω ≤ C  curl u0,Ω . Results on regular decompositions in this special case can be found in [24, 28].

3.2 Regular Potentials for Some Divergence-Free Functions Let the domain Ω and the Dirichlet boundary part ΓD be as introduced in Sect. 2 and let Γi , i = 0, . . . , β2 (Ω), denote the connected components of ∂Ω, where β2 (Ω) is the second Betti number of Ω. We define the space4   HΓD (div 00, Ω) := q ∈ HΓD (div 0, Ω) : γn q, 1Γi = 0, i = 0, . . . , β2 (Ω) . (9) Above γn denotes the normal trace operator, and the duality pairing is that between H −1/2(Γi ) and H 1/2(Γi ). If ΓD = ∅ we simply drop the subscript ΓD . Obviously, HΓD (div 00, Ω) ⊂ H(div 00, Ω) . The next result identifies the above space as the range of the curl operator. Theorem 3 (Regular Potential of Range(curl)) Let (Ω, ΓD ) be as in Sect. 2 and assume in addition that each connected component ΥD,k of the bulge has vanishing first Betti number, β1 (ΥD,k ) = 0. Then HΓD (div 00, Ω) = curl HΓD (curl, Ω) = curl H1ΓD (Ω) , and for each q ∈ HΓD (div 00, Ω) there exists ψ ∈ H1ΓD (Ω) depending linearly on q such that curl ψ = q

and  ∇ ψ0,Ω +

1 ψ0,Ω ≤ C q0,Ω , d(Ω)

where C depends only on the shape of Ω and ΓD , but not on d(Ω)

4 Alternatively

we can define HΓD (div 00, Ω) as the functions in HΓD (div 0, Ω) orthogonal to the harmonic Dirichlet fields H(div 0, Ω) ∩ H0 (curl 0, Ω).

Regular Decompositions

51

Remark 4 For the case that ΓD = ∅, we reproduce the classical result H(div 00, Ω) = curl H(curl, Ω) = curl H1 (Ω), see [11, Thm. 3.4]. In that case, Step 4 of the proof can be left out and ψ = w1 which is why div ψ = 0 in Ω. This property, however, is lost in the general case.

3.3 Rotation-Bounded Regular Decomposition of H(curl) We can now formulate another new variety of regular decompositions, for which the H1 -component will vanish for curl-free fields. Theorem 4 (Rotation-Bounded Regular Decomposition of H(curl) (I)) Let (Ω, ΓD ) be as in Sect. 2 and assume, in addition, that each connected component ΥD,k of the bulge has vanishing first Betti number, β1 (ΥD,k ) = 0. Then, for each u ∈ HΓD (curl, Ω) there exist z ∈ H1ΓD (Ω) and a curl-free vector field h ∈ HΓD (curl 0, Ω), depending linearly on u such that u = z + h, h0,Ω ≤ u0,Ω + C d(Ω)  curl u0,Ω ,  ∇ z0,Ω +

1 z0,Ω ≤ C  curl u0,Ω , d(Ω)

where C depends only on the shape of Ω and ΓD , but not on d(Ω). Remark 5 The constant C in Theorem 4 depends essentially on the stability div,0 and the (adapted) Stein constants of the divergence-free extension operator EΩ e . extension operator EΥ∇,Stein D Another stronger version of the rotation-bounded regular decomposition of H(curl) gets rid of the assumptions on the topology of the Dirichlet boundary and has improved stability properties (though with less explicit constants). Theorem 5 (Rotation-Bounded Regular Decomposition of H(curl) (II)) Let (Ω, ΓD ) be as in Sect. 2. Then for each u ∈ HΓD (curl, Ω) there exist z ∈ H1ΓD (Ω) and a curl-free h ∈ HΓD (curl 0, Ω) depending linearly on u such that u = z + h, z0,Ω + h0,Ω ≤ C u0,Ω ,  ∇ z0,Ω + d(Ω)−1 z0,Ω ≤ C  curl u0,Ω , where C depends only on the shape of Ω and ΓD , but not on d(Ω).

52

R. Hiptmair and C. Pechstein

Remark 6 For the case ΓD = ∂Ω the result of the theorem is already proved by 1 Remark 3 since we obtain u = z + ∇p with ∇p ∈ ∇H0,const (Ω) = H0 (curl, Ω). Remark 7 We would like to emphasize that both in Theorems 2 and 5, the domain Ω may be non-convex, non-smooth, and may have non-trivial topology: It may have handles and its boundary may have multiple components. Also the Dirichlet boundary ΓD may have multiple components, each of which with non-trivial topology. Moreover, we have the pure L2 (Ω)-stability in both theorems. In this sense, the results of Theorems 2 and 5 are superior to those found, e.g., in [7, Thm 3.4], [19] or the more recent ones in [8, Thm. 2.3], [22]. Remark 8 If Ω has vanishing first Betti number, β1 (Ω) = 0, then HΓD (curl 0, Ω) = ∇HΓ1D ,const (Ω). Hence, we can split each u ∈ HΓD (curl, Ω) into z ∈ H1ΓD (Ω) and ∇p with p ∈ H 1 (Ω) being constant on each connected component of ΓD . If ΓD is connected, then p ∈ HΓ1D (Ω). Summarizing, if Ω has no handles and if ΓD is connected, then we have the combined features of Theorems 2 and 5. Finally, we mention that the regular decomposition theorems spawn projection operators that play a fundamental role in the analysis of weak formulations of Maxwell’s equations in frequency domain [14, Sect. 5]. Corollary 1 Let (Ω, ΓD ) be as in Sect. 2. Then there exist continuous projection operators R : HΓD (curl, Ω) → H1ΓD (Ω) and N : HΓD (curl, Ω) → HΓD (curl 0, Ω) such that R + N = id and RvH1 (Ω) + NvL2 (Ω) ≤ C vH(curl,Ω)

∀v ∈ H(curl, Ω),

where C is a constant independent of v. Moreover, F : HΓD (curl, Ω) HΓD (curl, Ω) defined by Fv := Rv − Nv is an isomorphism.

→

Remark 9 The L2 -estimates from Theorem 4 then show that the corresponding operator R can be extended to a continuous operator mapping from L2 (Ω) to L2 (Ω).

4 Regular Decompositions and Potentials Related to H(div) The developments of this section are largely parallel to those of Sect. 3 with some new aspects concerning extensions and topological considerations.

4.1 Rotation-Based Regular Decomposition of H(div) The following theorem is the H(div)-counterpart of Theorem 2.

Regular Decompositions

53

Theorem 6 (Rotation-Based Regular Decomposition of H(div)) Let (Ω, ΓD ) satisfy the assumptions made in Sect. 2. Then for each v ∈ HΓD (div, Ω) there exist z ∈ H1ΓD (Ω) and q ∈ H1ΓD (Ω) depending linearly on v such that v = z + curl q, 1 q0,Ω ≤ C v0,Ω , d(Ω)   1 1 +  ∇ q0,Ω ≤ C  curl v0,Ω + v0,Ω , d(Ω) d(Ω)

z0,Ω +  curl q0,Ω +  ∇ z0,Ω +

1 z0,Ω d(Ω)

with constant C depending only on the shape of Ω and ΓD , but not on d(Ω).

4.2 Regular Potential with Prescribed Divergence The next result carries Theorem 3 over to H(div). Theorem 7 (Regular Potentials for the Image Space of div) Let (Ω, ΓD ) be as in Sect. 2 and, in addition, assume that each connected component ΥD,k of the bulge has a connected boundary, i.e., β2 (ΥD,k ) = 0. Then L2 (Ω) = div HΓD (div, Ω) = div H1ΓD (Ω). Moreover, for each v ∈ L2 (Ω) there exists q ∈ H1ΓD (Ω) depending linearly on v such that, with a constant C depending on Ω and ΓD but not on d(Ω), div q = v

and  ∇ q0,Ω +

1 q0,Ω ≤ C v0,Ω . d(Ω)

4.3 Divergence-Bounded Regular Decompositions of H(div) We can now formulate other variants of regular decompositions of H(div) in analogy to what we did in Sect. 3.3. Theorem 8 (Divergence-Bounded Regular Decomposition of H(div) (I)) Let (Ω, ΓD ) be as in Sect. 2. In addition, assume that each connected component ΥD,k of the bulge has a connected boundary, i.e., β2 (ΥD,k ) = 0. Then, for each v ∈ HΓD (div, Ω) there exists z ∈ H1ΓD (Ω) and a divergence-free vector field

54

R. Hiptmair and C. Pechstein

h ∈ HΓD (div 0, Ω) depending linearly on v such that v = z + h,

(10)

h0,Ω ≤ v0,Ω + C d(Ω) div v0,Ω ,  ∇ z0,Ω +

1 z0,Ω ≤ C  div v0,Ω , d(Ω)

(11) (12)

where C depends only on the shape of Ω and ΓD , but not on d(Ω). The last variant of H(div) regular decomposition of H(div) dispenses with the assumptions on the topology of the Dirichlet boundary and has better stability properties than the splitting from Theorem 8 (though with less explicit constants). Theorem 9 (Divergence-Bounded Regular Decomposition of H(div) (II)) Let (Ω, ΓD ) be as in Sect. 2. Then, for each v ∈ HΓD (div, Ω) there exists z ∈ H1ΓD (Ω) and a divergence-free vector field h ∈ HΓD (div 0, Ω) depending linearly on v such that v = z + h, z0,Ω + h0,Ω ≤ v0,Ω ,  ∇ z0,Ω +

1 z0,Ω ≤ C  div v0,Ω , d(Ω)

(13) (14) (15)

where C depends only on the shape of Ω and ΓD , but not on d(Ω).

5 Discrete Counterparts of the Regular Decompositions The discrete setting to which we want to extend the concept of regular decompositions is provided by finite element exterior calculus (FEEC, [2]) which introduces finite element subspaces of H(curl) and H(div) as special instances of spaces of discrete differential forms. In this section we confine ourselves to the lowest-order case of piecewise linear finite element functions. Throughout, we assume that (Ω, ΓD ) is as in Sect. 2, and, additionally, that Ω is a polyhedron and that ∂ΓD consists of straight line segments. All considerations take for granted a shape-regular family of meshes {T h }h of Ω, consisting of tetrahedral elements, and resolving ΓD in the sense that ΓD is a union of faces of some of the tetrahedra. The following finite element spaces will be relevant: 0 • the space Wh,Γ (Ω) of HΓ1D (Ω)-conforming piecewise linear Lagrangian finite D element functions, • the space W 1h,ΓD (Ω) of HΓD (curl, Ω)-conforming lowest order Nédélec elements, also known as edge elements,

Regular Decompositions

55

• the space W 2h,ΓD (Ω) of HΓD (div, Ω)-conforming lowest order tetrahedral Raviart-Thomas finite elements, aka, face elements, 0 • the space W 0h,ΓD (Ω) := [Wh,Γ (Ω)]3 of piecewise linear globally continuous D vector fields vanishing on ΓD . Functions in W h,ΓD (Ω),  = 1, 2, 3, are so-called Whitney forms, lowest-order discrete differential forms of the first family as introduced in [13] and [2, Sect. 5].

5.1 Discrete Regular Decompositions for Edge Elements Commuting projectors, also known as co-chain projectors, are the linchpin of FEEC theory [2, Sect. 7], and it is not different with our developments. Thus, let 0 0 Rh,Γ : HΓ1D (Ω) → Wh,Γ (Ω) D D 1 and R1h,ΓD : HΓD (curl, Ω) → Wh,Γ (Ω) D

denote the continuous, boundary-aware cochain projectors from [17, Sect. 3.2.6], which extend the pioneering work [10] by Falk and Winther. These two linear operators are projectors onto their ranges, they fulfill the commuting property 0 ϕ) = R1h,ΓD (∇ ϕ) ∇(Rh,Γ D

∀ϕ ∈ HΓ1D (Ω) ,

(16)

and local stability estimates Theorem 10 ([17, Thm. 1.2]) For each vh ∈ W 10,ΓD (Ω) there exists a continuous and piecewise linear vector field zh ∈ W 0h,ΓD (Ω), a continuous and piecewise 0 (Ω), and a remainder 3 vh ∈ W 10,ΓD (Ω), all linear scalar function ph ∈ Wh,Γ D depending linearly on vh , providing the discrete regular decomposition vh = R1h,ΓD zh +3 vh + ∇ ph and satisfying the stability estimates zh 0,Ω +  ∇ ph 0,Ω + 3 vh 0,Ω ≤ C vh 0,Ω ,   ∇ zh 0,Ω + h−13 vh 0,Ω ≤ C  curl vh 0,Ω +

(17) 1 d(Ω) vh 0,Ω



,

(18)

where C is a generic constant that depends only on the shape of (Ω, ΓD ), but not on d(Ω), and on the shape regularity constant of T h (Ω). Above, h−1 is the piecewise constant function that is equal to h−1 T on every element T . Obviously, this is a discrete counterpart of the regular decomposition of H(curl) from Theorem 2. The following theorem appears to be new and it corresponds to

56

R. Hiptmair and C. Pechstein

the rotation-bounded regular decomposition of Theorem 5. For the sake of brevity define the discrete nullspace of the curl operator Nh1 := {vh ∈ W 1h,ΓD (Ω) : curl vh = 0} .

(19)

0 (Ω), but if the first Betti number If Ω and ΓD have simple topology, Xh = ∇ Wh,Γ D of Ω is non-zero, or if ΓD has multiple components, then a finite-dimensional cohomology space has to be added [2, Sect. 5.6].

Theorem 11 (Rotation-Bounded Discrete Regular Decomposition for Edge Elements) For each vh ∈ W 10,ΓD (Ω) there exists a continuous and piecewise linear vector field zh ∈ W 0h,ΓD (Ω), an curl-free edge element function hh ∈ Nh1 , and a remainder 3 vh ∈ W 10,ΓD (Ω), all depending linearly on vh , providing the discrete regular decomposition vh = R1h,ΓD zh +3 vh + hh and satisfying the stability bounds ⎫ zh 0,Ω ⎪ ⎬ hh 0,Ω ≤ C vh 0,Ω , ⎪ 3 vh 0,Ω ⎭

 ∇ zh 0,Ω h−13 vh 0,Ω

7 ≤ C  curl vh 0,Ω ,

where C is a uniform constant that depends only on the shape of (Ω, ΓD ), but not on d(Ω), and on the shape regularity constant of T h (Ω). We stress that the statements of Theorems 10 and 11 do not hinge on any assumptions on the topological properties of Ω and ΓD .

5.2 Discrete Regular Decompositions for Face Elements For face elements, the construction of a boundary-aware co-chain projection operator R2h,ΓD : HΓD (div, Ω) → W 2h,ΓD (Ω) that commutes with R1h,ΓD and the curl-operator has not yet been accomplished. Fortunately, in the case ΓD = ∅, this operator is available from [10]. Thus, in the following, we treat only the case ΓD = ∅ and just omit the subscript ΓD . Then, from [10] we can borrow a linear operator R2h : H(div, Ω) → W 2h (Ω) such that curl R1h u = R2h curl u

∀u ∈ H(curl, Ω) .

The next result takes Theorem 6 to the discrete setting.

(20)

Regular Decompositions

57

Theorem 12 (Discrete Regular Decomposition of W 2h (Ω)) For each vector field vh in the lowest-order Raviart-Thomas space W 2h (Ω), there exists a continuous and piecewise linear vector field zh ∈ W 0h (Ω), a vector field qh in the lowest-order vh ∈ W 2h (Ω), all depending linearly on Nédélec space W 1h (Ω), and a remainder 3 vh , providing the discrete regular decomposition vh = R2h zh +3 vh + curl qh , and the stability estimates

 curl qh 0,Ω

⎫ zh 0,Ω ⎪ ⎬ 1 + d(Ω) qh 0,Ω ≤ C vh 0,Ω , ⎪ 3 vh 0,Ω ⎭ 7 1  ∇ zh 0,Ω ≤ C div vh 0,Ω + vh 0,Ω . h−13 vh 0,Ω d(Ω)

The constant C depends only on the shape of Ω, but not on d(Ω), and the shaperegularity of T h (Ω). Finally, we present a counterpart to the divergence-bounded regular decomposition of Theorem 9. For convenience we introduce the space of divergence-free face element functions Nh2 := {qh ∈ W 2h (Ω) : div qh = 0} .

(21)

Theorem 13 (Divergence-Bounded Discrete Regular Decomposition of W 2h (Ω)) For each vector field vh in the lowest-order Raviart-Thomas space W 2h (Ω), there exists a continuous and piecewise linear vector field zh ∈ W 0h (Ω), an element hh in the discrete divergence-free subspace Nh2 , and a remainder 3 vh ∈ W 2h (Ω), all depending linearly on vh , providing the discrete regular decomposition vh + hh vh = R2h zh +3 and the stability estimates ⎫ zh 0,Ω ⎪ ⎬ 3 vh 0,Ω ≤ C vh 0,Ω , ⎪ hh 0,Ω ⎭

 ∇ zh 0,Ω h−13 vh 0,Ω

7 ≤ C  div vh 0,Ω .

The constants C depend only on the shape of Ω, but not on d(Ω), and the shape regularity of T h (Ω).

58

R. Hiptmair and C. Pechstein

Remark 10 The result of Theorem 13 can be viewed as an improvement of the decompositions in [25] which are elaborated for the case of essential boundary conditions on ∂Ω. Corollary 2 If the second Betti number of Ω vanishes, that is, if ∂Ω is connected, then hh in Theorem 13 can be chosen as hh = curl qh with qh ∈ W 1h (Ω) such that vh + curl qh , vh = R2h z +3 with the bounds ⎫ zh 0,Ω ⎪ ⎪ ⎪ ⎬ 3 vh 0,Ω ≤ C vh 0,Ω ,  curl qh 0,Ω ⎪ ⎪ ⎪ d(Ω)−1 qh 0,Ω ⎭

 ∇ zh 0,Ω h−13 vh 0,Ω

7 ≤ C  div vh 0,Ω .

Remark 11 The result of Corollary 2 is an improvement of [19, Lemma 5.2] which assumes a domain Ω that is smooth enough to allow H 2 -regularity of the Laplace problem (2-regular case, for details see [19, Sect. 3]). This lemma is used in [27] in a domain decomposition framework, where convex subdomains are assumed. With our improved version, this assumption can be weakened considerably. Acknowledgements The second author would like to thank Dirk Pauly (Essen) for enlightening discussions.

References 1. Amrouche, C., Bernardi, C., Dauge, M., Girault, V.: Vector potentials in three-dimensional non-smooth domains. Math. Meth. Appl. Sci. 21(9), 823–864 (1998) 2. Arnold, D., Falk, R., Winther, R.: Finite element exterior calculus, homological techniques, and applications. Acta Numer. 15, 1–155 (2006) 3. Bauer, S., Pauly, D., Schomburg, M.: The Maxwell compactness property in bounded weak Lipschitz domains with mixed boundary conditions. SIAM J. Math. Anal. 48(4), 2912–2943 (2016) 4. Birman, M.S., Solomyak, M.Z.: L2 -theory of the Maxwell operator in arbitrary domains. Usp. Mat. Nauk. 42(6), 61–76 (1987). English Transliterated, Russian Math. Surveys 42(6), 75–96 (1987) 5. Bonnet-Ben Dhia, A., Hazard, C., Lohrengel, S.: A singular field method for the solution of Maxwell’s equations in polyhedral domains. SIAM J. Appl. Math. 59(6), 2028–2044 (1999) 6. Buffa, A., Costabel, M., Sheen, D.: On traces for H(curl, Ω) in Lipschitz domains. J. Math. Anal. Appl. 276, 845–867 (2002) 7. Costabel, M., Dauge, M., Nicaise, S.: Singularities of Maxwell interface problems. Model. Math. Anal. Numer. 33(2), 627–649 (1999) 8. Creusé, E., Nicaise, S., Tang, Z.: Helmholtz decomposition of vector fields with mixed boundary conditions and an application to a posteriori finite element analysis of the Maxwell system. Math. Methods Appl. Sci. 38(4), 738–750 (2015)

Regular Decompositions

59

9. Dautray, R., Lions, J.: Mathematical Analysis and Numerical Methods for Science and Technology. Springer, Berlin (2000). Original French edition published by Masson, S.A., Paris (1984) 10. Falk, R.S., Winther, R.: Local bounded cochain projections. Math. Comp. 83(290), 2631–2656 (2014) 11. Girault, V., Raviart, P.A.: Finite Element Methods for Navier–Stokes Equations. Springer, New York (1986) 12. Gopalakrishnan, J., Qiu, W.: Partial expansion of a Lipschitz domain and some applications. Front. Math. China 7(2), 249–272 (2012) 13. Hiptmair, R.: Canonical construction of finite elements. Math. Comp. 68, 1325–1346 (1999) 14. Hiptmair, R.: Finite elements in computational electromagnetism. Acta Numer. 11, 237–339 (2002) 15. Hiptmair, R.: Boundary element methods for eddy current computation. In: Schanz, M., Steinbach, O. (eds.) Boundary Element Analysis: Mathematical Aspects and Applications. Lecture Notes in Applied and Computational Mechanics, vol. 29, pp. 213–248. Springer, Heidelberg (2007) 16. Hiptmair, R.: Maxwell’s equations: Continuous and discrete. In: Bermúdez de Castro, A., Valli, A. (eds.) Computational Electromagnetism. Lecture Notes in Mathematics, vol. 2148. Springer, Cham (2015) 17. Hiptmair, R., Pechstein, C.: Discrete regular decompositions of tetrahedral discrete 1-forms. In: Langer, U., Pauly, D., Repin, S. (eds.) Maxwell’s Equations. Radon Series on Computational and Applied Mathematics, vol. 24, pp. 199–258. De Gruyter, Stuttgart (2019) 18. Hiptmair, R., Pechstein, C.: Regular decompositions of vector fields - continuous, discrete, and structure-preserving. Technical Report 2019-18, Seminar for Applied Mathematics, ETH Zürich (2019). https://www.sam.math.ethz.ch/sam_reports/reports_final/reports2019/2019-18. pdf 19. Hiptmair, R., Xu, J.: Nodal auxiliary space preconditioning in H(curl) and H(div) spaces. SIAM J. Numer. Anal. 45, 2483–2509 (2007) 20. Hiptmair, R., Zheng, W.: Local multigrid in H(curl). J. Comput. Math. 27(5), 573–603 (2009) 21. Hiptmair, R., Widmer, G., Zou, J.: Auxiliary space preconditioning in H0 (curl; Ω). Numer. Math. 103, 435–459 (2006) 22. Hu, Q.: Convergence of HX preconditioner for Maxwell’s equations with jump coefficients (i): various extensions of the regular Helmholtz decomposition. Technical Report arXiv:1708.05850v2 [math.AP] (2018) 23. Kolev, T.V., Pasciak, J.E., Vassilevski, P.S.: H(curl) auxiliary mesh preconditioning. Numer. Linear Algebr. Appl. 15(5), 455–471 (2008) 24. Kolev, T.V., Vassilevski, P.S.: Parallel auxiliary space AMG for H(curl) problems. J. Comput. Math. 27(5), 604–623 (2009) 25. Kolev, T.V., Vassilevski, P.S.: Parallel auxiliary space AMG for H(div) problems. SIAM J. Sci. Comput. 34(6), A3079–A3098 (2012) 26. Monk, P.: Finite Element Methods for Maxwell’s Equations. Numerical Mathematics and Scientific Computation. Oxford University Press, Oxford (2003) 27. Oh, D., Widlund, O.B., Zampini, S., Dohrmann, C.R.: BDDC algorithms with deluxe scaling and adaptive selection of primal constraints for Raviart–Thomas vector fields. Math. Comp. 87, 659–692 (2018) 28. Pasciak, J.E., Zhao, J.: Overlapping Schwarz methods in H(curl) on polyhedral domains. Numer. Math. 10(3), 211–234 (2002)

60

R. Hiptmair and C. Pechstein

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Model Reduction by Separation of Variables: A Comparison Between Hierarchical Model Reduction and Proper Generalized Decomposition Simona Perotto, Michele Giuliano Carlino, and Francesco Ballarin

1 Introduction This paper is meant as a first attempt to compare two procedures which share the idea of exploiting separation of variables to perform model reduction, albeit with different purposes. Proper Generalized Decomposition (PGD) is essentially employed as a powerful tool to deal with parametric problems in several fields of application [3, 14, 23]. Parametrized models characterize multi-query contexts, such as parameter optimization, statistical analysis or inverse problems. Here, the computation of the solution for many different parameters demands, in general, a huge computational effort, and this justifies the development of model reduction techniques. For this purpose, projection-based techniques, such as Proper Orthogonal Decomposition (POD) or Reduced Basis methods, are widely used in the literature [11]. The idea is to project the discrete operators onto a reduced space so that the problem can be solved rapidly in the lower dimensional space. PGD adopts a completely different way to deal with parameters. Here, parameters are considered

S. Perotto MOX - Modeling and Scientific Computing, Dipartimento di Matematica, Politecnico di Milano, Milano, Italy e-mail: [email protected] M. G. Carlino Inria Bordeaux Sud-Ouest and Institut de Mathématiques de Bordeaux, University of Bordeaux, Talence, France e-mail: [email protected] F. Ballarin () MathLab, Mathematics Area, SISSA, Trieste, Italy e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_4

61

62

S. Perotto et al.

as new independent variables of the problem, together with the standard space-time ones [5]. Although the dimensionality of the problem is inevitably increased, PGD transforms the computation of the solution for new values of the parameters into a plain evaluation of the reduced solution, with striking computational advantages. Hierarchical-Model (HiMod) reduction has been proposed to improve onedimensional (1D) partial differential equation (PDE) solvers for problems defined in domains with a geometrically dominant direction, like slabs or pipes [6, 20]. The main applicative field of interest is hemodynamics, in particular the modeling of blood flow in patient-specific geometries. Purely 1D hemodynamic models completely drop the transverse dynamics, which, however may be locally important (e.g., in the presence of a stenosis or an aneurism). HiMod aims at providing a numerical tool to incorporate the transverse components of the 3D solution into a conceptually 1D solver. To do this, the driving idea is to discretize main and transverse dynamics in a different way. The latter are generally of secondary importance and can be described by few degrees of freedom using a spectral approximation, in combination, for instance, with a finite element (FE) discretization of the mainstream. The parametric version of HiMod (namely, HiPOD) is a more recent proposal [4, 13]. On the other hand, PGD is not so widely employed in a non-parametric setting, despite its original formulation [12]. Nevertheless, for the sake of comparison, in this paper we consider the non-parametric as well as the parametric versions of both the HiMod and PGD approaches. The goal is to begin a preliminary comparative analysis between the two methodologies, to highlight the respective weaknesses and strengths. The main limit of PGD remains its inability to deal with non-Cartesian geometries without losing the computational benefits arising from the separability of the spatial coordinates. HiMod turns out to be more flexible from a geometric viewpoint. On the other hand, PGD turns out to be extremely effective for parametric problems thanks to the explicit expression of the PGD solution in terms of the parameters, while HiPOD can be classified as a projection-based method with all the associated drawbacks. In perspective, the ultimate goal is to merge HiMod with PGD to emphasize the good features and mitigate the intrinsic limits of the two methods taken alone.

2 The HiMod Approach Hierarchical Model reduction proved to be an efficient and reliable method to deal with phenomena characterized by dominant dynamics [10]. In general, the computational domain itself exhibits an intrinsic directionality. We8assume  ⊂ Rd (d = 2, 3) to coincide with a d-dimensional fiber bundle,  = x∈1D {x} × γx , where 1D ⊂ R denotes the supporting fiber aligned with the main stream, while γx ⊂ Rd−1 is the transverse fiber at x ∈ 1D , parallel to the transverse dynamics. For the sake of simplicity, we identify 1D with a straight segment, (x0 , x1 ). We refer to [15, 21] for the case where 1D is curvilinear. From a computational

Model Reduction by Separation of Variables: A Comparison Between HiMod and PGD

63

ˆ transforming the physical viewpoint, the idea is to exploit a map,  :  → , ˆ and to make explicit computations in  ˆ only. domain, , into a reference domain, , ˆ coincides with a rectangle in 2D, with a cylinder with circular section Typically,  in 3D. To define , for each x ∈ 1D , we introduce the map, ψx : γx → γˆd−1 , from fiber γx to the 8 reference transverse fiber, γˆd−1 , so that the reference domain ˆ = x∈ {x} × γˆd−1 . The supporting fiber is preserved by map coincides with  1D , which modifies the lateral boundaries only. We consider now the (full) problem to be reduced. Due to the comparative purposes of the paper, we focus on a scalar elliptic equation, and, in particular, on the associated weak formulation, find u ∈ V : a(u, v) = F (v)

∀v ∈ V ,

(1)

where V ⊆ H 1 (), a(·, ·) : V × V → R is a continuous and coercive bilinear form and F (·) : V → R is a continuous linear functional. To provide the HiMod formulation for problem (1), we introduce the hierarchical reduced space + Vm = vm (x, y) =

m 

7 h v˜k (x)ϕk (ψx (y)), with v˜k ∈ V1D , x ∈ 1D , y ∈ γx

k=1

(2) h ⊆ H 1 ( ) is a discrete space of dimension for a modal index m ∈ N+ , where V1D 1D Nh associated with a partition Th of 1D , while {ϕk }m k=1 denotes a modal basis of functions orthogonal with respect to the L2 (γˆd−1 )-scalar product. Index m sets the hierarchical level of the HiMod space, being Vm ⊂ Vm+1 , for any m. Concerning h V1D , we adopt here a standard FE space, although any discrete space can be employed (see, e.g., [21], where an isogeometric discretization is used). Functions h in V1D have to include the boundary conditions on {x0 } × γx0 and {x1} × γx1 ; analogously, the modal functions have to take into account the boundary data along the horizontal sides. In Sect. 4 further comments are provided about the selection of the modal basis and of the modal index m. The HiMod formulation for problem (1) thus reads

∈ Vm : a(uHiMod , vm ) = F (vm ) find uHiMod m m

∀vm ∈ Vm .

(3)

To ensure the well-posedness of formulation (3) and the convergence of the HiMod approximation, uHiMod , to the full solution, u, we endow the HiMod space with a m conformity and a spectral approximability hypothesis, and we introduce a standard h density assumption on the discrete space V1D (see [20] for all the details). h The HiMod solution can be fully characterized by introducing a basis, {θl }N l=1 , h HiMod for the space V1D . Actually, each modal coefficient, u˜ k , of um can be expanded

64

S. Perotto et al.

in terms of such a basis, so that, we obtain the modal representation uHiMod (x, y) = m

Nh m  

u˜ k,l θl (x)ϕk (ψx (y)).

(4)

k=1 l=1 h The actual unknowns of problem (3) become the mNh coefficients {u˜ k,l }m,N k=1,l=1 . With reference to the Poisson problem, − u = f , completed with full homogeneous Dirichlet boundary data, the corresponding HiMod formulation, after exploiting (4) in (3) and picking vm (x, y) = θi (x)ϕj (ψx (y)) with i = 1, . . . , Nh and j = 1, . . . , m, reduces to the system of mNh 1D equations in the mNh h unknowns {u˜ k,l }m,N k=1,l=1 ,

Nh m   k=1 l=1

9

u˜ k,l



dθl dθl dθi (x) (x) + rˆj1,0 k (x) dx (x)θi (x)+ dx dx 1D

:  dθi 0,1 0,0 fˆj (x)θi (x) dx, (x) + rˆj k (x)θl (x)θi (x) dx = + rˆj k (x)θl (x) dx 1D rˆj1,1 k (x)

;   a,b −1 y)) ˆ )|J | d yˆ with a, b = 0,1, J = det D−1 where rˆja,b k (x) = γˆd−1 rj k (x, y 2 (x, ψx (ˆ with D2 = D2 (x, ψx−1 (ˆy)) = ∇y ψx ,   ˆ ) = ϕk (ˆy)ϕj (ˆy) D21 + D22 , rj0,1 ˆ ) = ϕk (ˆy)ϕj (ˆy)D1 , rj0,0 k (x, y k (x, y ˆ ) = ϕk (ˆy)ϕj (ˆy)D1 , rj1,0 k (x, y

ˆ ) = ϕk (ˆy)ϕj (ˆy), rj1,1 k (x, y

; with D1 = D1 (x, ψx−1 (ˆy)) = ∂ψx /∂x, and fˆj (x) = γˆd−1 f (x, ψx−1 (ˆy))ϕj (ˆy) |J | d yˆ . Information associated with the transverse dynamics are lumped in the coefficients {ˆrja,b k }, so that the HiMod system is solved on the supporting fiber, 1D . Collecting the HiMod unknowns, by mode, in the vector uHiMod ∈ RmNh , such that m = [u˜ 1,1 , u˜ 1,2 , . . . , u˜ 1,Nh , u˜ 2,1 , . . . , u˜ m,1 , . . . , u˜ m,Nh ]T , uHiMod m

(5)

we can rewrite the HiMod system in the compact form uHiMod = fHiMod , AHiMod m m m

(6)

where AHiMod ∈ RmNh ×mNh and fHiMod ∈ RmNh are the HiMod stiffness m m ; matrix and right-hand side, respectively, with [fHiMod ]j i = 1D fˆj (x)θi (x)dx, and m ; < d a θl d b θi [AHiMod ]j i,kl = 1a,b=0 1D rˆja,b m k (x) dx (x) dx (x)dx. According to (5), for each modal index j , between 1 and m, the nodal index, i, takes the values 1, . . . , Nh . Thus, HiMod reduction leads to solve a system of order mNh , independently of the dimension of the full problem (1).

Model Reduction by Separation of Variables: A Comparison Between HiMod and PGD

65

3 The PGD Approach To perform PGD, we have to introduce on problem (1) a separability hypothesis with respect to both the spatial variables and the data [5, 22]. Thus, domain  ⊂ Rd coincides with the rectangle x × y if d = 2, with the parallelepiped x × y × z (total separability) or with the cylinder x × y (partial separability) if d = 3, for x , y , z ⊂ R and y ⊂ R2 , being y = (y, z). In the following, we focus on partial separability, since it is more suited to match HiMod reduction with PGD. Analogously, we assume that the generic problem data, d = d(x, y, z), can be written as d = d x (x)d y (y). The separability is inherited by the PGD space + Wm = wm (x, y) =

m 

7 y y y wkx (x)wk (y), with wkx ∈ Whx , wk ∈ Wh , x ∈ x , y ∈ y ,

k=1

(7) y

where Whx ⊆ H 1 (x ) and Wh ⊆ H 1 (y ; Rd−1) are discrete spaces, with y y y dim(Whx ) = Nhx and dim(Wh ) = Nh , associated with partitions, Txh and Th , of y x x and y , respectively. In general, Wh and Wh are FE spaces, although, a priori, any discretization can be adopted. It turns out that Wm is a tensor function space, y being Wm = Whx ⊗ Wh ⊆ H 1 (x ) ⊗ H 1 (y ; Rd−1). Index m plays the same role as in the HiMod reduction, setting the level of detail for the reduced solution (see Sect. 4 for possible criteria to choose m). PGD exploits the hierarchical structure in Wm to build the generic function wm ∈ Wm . In particular, wm is computed as x y wm (x, y) = wm (x)wm (y) +

m−1 

y

wkx (x)wk (y),

(8)

k=1 y

where wkx and wk are assumed known for k = 1, . . . , m − 1, so that the x and w y , become the actual unknowns. To provide enrichment functions, wm m the PGD formulation for the Poisson problem considered in Sect. 2, we exploit representation (8) for the PGD approximation, uPGD m , and we pick the test function y as X(x)Y (y), with X ∈ Whx and Y ∈ Wh . The coupling between the unknowns, uxm y and um , leads to a nonlinear problem, which is tackled by means of the Alternating y Direction Strategy (ADS) [5]. The idea is to look for uxm and um , separately via a fixed point procedure. We introduce an auxiliary index to keep trace of the ADS x,p y,p iterations, so that, at the p-th ADS iteration we compute um and um starting x,g y,g from the previous approximations, um and um for g = 1, . . . , p − 1, following a x,p y,p−1 y , and by two-step procedure. First, we compute um by identifying um with um

66

S. Perotto et al. y,p−1

selecting Y (y) = um

in the test function. This yields, for any X ∈ Whx ,

; ;  y,p−1  2  x,p   ;  y,p−1 2 x,p dy + x um Xdx y um dy um X dx y um ;   ; ; ; < y,p−1 y y,p−1 x   dy − m−1 dy = x f x Xdx y f y um k=1 x uk X dx y uk um     ; 0. For details on the method implementation and its convergence properties we refer to [7].

4 Numerical Examples We test now our method on three different examples. In each of them, we specify the setting and the parameters used to build the surrogate and visualize our approximation to the center manifold. Additionally, we compute the pointwise residual     r(x) = Dsh (x) L1 x + N1 (x, sh (x)) − L2 sh (x) + N2 (x, sh (x)) , which measures how well the surrogate sh satisfies the ODE (4). In all the three examples, the greedy algorithm is used to select a suitable subset of the points, and in all cases the procedure is stopped with a prescribed εt ol . In the first two examples we set εt ol := 10−15 , while εt ol := 10−10 is used in the last one.

4.1 Example 1 We consider the 2-dimensional system x˙ = L1 x + N1 (x, y) = 0 + xy y˙ = L2 y + N2 (x, y) = −y + x 2 .

(10)

We generate the training data by solving (10) with an implicit Euler scheme for initial time t0 = 0, final time T = 1000 and with the time step Δt = 0.1. We initiate the numerical procedure with initial values (x0 , y0 ) ∈ {±0.8} × {±0.8} and store the resulting data pairs in X and Y after discarding all data whose x-values are not contained in the neighborhood [−0.1, 0.1] which results in N ∗ = 38,248 data pairs.  4 We run the greedy algorithm for the kernels k1 (x, y) := 1 + xy/2 and 2 k2 (x, y) = e−(x−y) /2 . This results in the sets X1 and X2 which contain 14 and 6 points, respectively. The corresponding approximations s1 and s2 for the constant

102

B. Haasdonk et al. ·10−2

Fig. 1 Approximations s1 and s2 of the center manifold

s1 s2

0

−0.5

−1 −0.1

Fig. 2 Residuals r1 and r2 of the center manifold

4

−0.05

0

0.05

0.1

·10−5 r1 r2

3 2 1 0 −0.1

−0.05

0

0.05

0.1

regularization function r ≡ 10−10 are plotted in Fig. 1 over the domain [−0.1, 0.1]. The pointwise residual is depicted in Fig. 2.

4.2 Example 2 We consider the 2-dimensional system x˙ = L1 x + N1 (x, y) = 0 − xy y˙ = L2 y + N2 (x, y) = −y + x 2 − 2y 2.

(11)

The training data is generated the same way as in Example 1. We again use the kernels k1 and k2 . The greedy algorithm gives sets X1 and X2 of size 12 and 6, respectively. The evaluation of the approximations s1 and s2 over the neighborhood

Greedy Kernel Methods for Center Manifold Approximation

103

·10−2

Fig. 3 Approximations s1 and s2 of the center manifold

s1 s2

1

0.5

0 −0.1

−0.05

0

0.05

0.1

·10−7

Fig. 4 Residuals r1 and r2 of the center manifold

r1 r2

1

0.5

0 −0.1

−0.05

0

0.05

0.1

[−0.1, 0.1] can be seen in Fig. 3, while the respective pointwise residuals are plotted in Fig. 4.

4.3 Example 3 We consider the (2 + 1)-dimensional system #

0 −1 x˙ = L1 x + N1 (x, y) = 1 0 y˙ = L2 y + N2 (x, y) = −y

$# $ # $ x1 x y + 1 x2 x2 y

− x12

− x22

(12)

+y . 2

We generate the training data in a similar fashion as before. We again use the implicit Euler scheme with start time t = 0, final time T = 1000 and with time step

104

B. Haasdonk et al.

Approximation s2

Approximation s1 ·10−2

·10−3

0

0

−0.5 −5

−1 −0.1

0.1 x1

0

0 0.1 −0.1

−0.1

x2

0.1 x1

0

0 0.1 −0.1

Residual r1

x2

Residual r2

·10−3

·10−2

1

5

0

0

−0.1

0.1 x1

0

0 0.1 −0.1

x2

−0.1

0.1 x1

0

0 0.1 −0.1

x2

Fig. 5 Approximations s1 and s2 of the center manifold and corresponding residuals r1 and r2

Δt = 0.1. The Euler method is performed for initial data (x0 , y0 ) ∈ {±0.8}3 and the resulting trajectories are stored in X and Y , where only data with x ∈ [−0.1, 0.1]2 was considered; this leads to N ∗ = 78,796 data pairs. We use the kernels k1 (x, y) = 2 (1 + x T y/2)4 and k2 (x, y) = e−x−y2 /2 , and the greedy-selected sets have the size 21 (for k1 ) and 25 (for k2 ), respectively. The approximations s1 , s2 and their corresponding residuals r1 and r2 computed over the domain [−0.1, 0.1]2. The results can be seen in Fig. 5. We remark that in all the three experiments both kernels give comparable results in terms of error magnitude, and they both provide a good approximation of the manifold.

Greedy Kernel Methods for Center Manifold Approximation

105

5 Conclusions In this paper we introduced a novel algorithm to approximate the center manifold of a given ODE using a data-based surrogate. This algorithm computes an approximation of the manifold from a set of numerical trajectories with different initial data. It is based on kernel methods, which allow the use of the scattered data generated by these simulations as training points. Moreover, an application-specific ansatz and cost function have been employed in order to enforce suitable properties on the surrogate. Several numerical experiments suggested that the present method can reach a significant accuracy, and that it has the potential to be used as an effective model reduction technique. It seems promising to apply this approach to high dimensional systems as the approximation technique straightforwardly can be extended and is less prone to the curse of dimensionality than grid-based approximation techniques. An interesting extension would consist of determining the decomposition (2) in a data-based fashion by suitable processing of the trajectory data. Acknowledgements The first, third, and fourth authors would like to thank the German Research Foundation (DFG) for support within the Cluster of Excellence in Simulation Technology (EXC 310/2) at the University of Stuttgart. The second author thanks the European Commission for financial support received through Marie Curie Fellowships.

References 1. Carr, J.: Applications of Centre Manifold Theory. Applied Mathematical Sciences, vol. 35. Springer, Berlin (1981) 2. De Marchi, S., Schaback, R., Wendland, H.: Near-optimal data-independent point locations for radial basis function interpolation. Adv. Comput. Math. 23(3), 317–330 (2005) 3. Henry, D.: Geometric Theory of Semilinear Parabolic Equations. Lecture Notes in Mathematics, vol. 840. Springer, Berlin (1981) 4. Kelley, A.: The stable, center-stable, center, center-unstable, unstable manifolds. J. Differ. Equ. 3, 546–570 (1967) 5. Micchelli, C.A., Pontil, M.: On learning vector-valued functions. Neural Comput. 17(1), 177– 204 (2005) 6. Pliss, V.A.: A reduction principle in the theory of stability of motion. Izv. Akad. Nauk SSSR Ser. Mat. 28, 1297–1324 (1964) 7. Santin, G., Haasdonk, B.: Convergence rate of the data-independent P-greedy algorithm in kernel-based approximation. Dolomites Res. Notes Approx. 10, 68–78 (2017) 8. Shoshitaishvili, A.N.: Bifurcations of topological type of singular points of vector fields that depend on parameters. Funkcional. Anal. i Priložen. 6(2), 97–98 (1972) 9. Wendland, H.: Scattered Data Approximation. Cambridge Monographs on Applied and Computational Mathematics, vol. 17. Cambridge University Press, Cambridge (2005) 10. Wittwar, D., Santin, G., Haasdonk, B.: Interpolation with uncoupled separable matrix-valued kernels. Dolomites Res. Notes Approx. 11, 23–29 (2018) 11. Wittwar, D., Santin, G., Haasdonk, B.: Weighted regularized interpolation with matrix valued kernels. Technical Report, University of Stuttgart, (2019, in preparation)

106

B. Haasdonk et al.

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

An Adaptive Error Inhibiting Block One-Step Method for Ordinary Differential Equations Jiaxi Gu and Jae-Hun Jung

1 Introduction General linear methods have been extensively studied for solving ODEs. Among the large family of general linear methods the diagonally implicit multistage integration methods (DIMSIMs) in [1] are the special cases, which exhibit considerable potential for efficient implementation, providing the global error of the same order as the local truncation error. In [2], it was demonstrated that finite difference methods for PDEs can be constructed such that their convergence rates, or the order of their global errors, are higher than the order of the truncation errors. Following this idea, Ditkowski and Gottlieb devised the error inhibiting strategy in [3] by inhibiting the lowest order term in the truncation error from accumulating over time and thus showed that the global error of the scheme is one order higher than the local truncation error. The form of the error inhibiting scheme is inspired by the work of [7], where a block of s new step values is obtained at each step. The key idea of this method is to construct a coefficient matrix that has the null space where the local truncation error resides. In this work, we further improved the original error inhibiting method by introducing an additional free parameter used in the radial basis function

J. Gu Department of Mathematics, University at Buffalo, The State University of New York, Buffalo, NY, USA e-mail: [email protected] J.-H. Jung () Department of Mathematics, University at Buffalo, The State University of New York, Buffalo, NY, USA Department of Data Science, Ajou University, Suwon, South Korea e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_7

107

108

J. Gu and J.-H. Jung

(RBF) approximations. The main idea of the proposed method is to adopt the free parameter in the reconstruction of the error inhibiting method and to control it for further possible error cancellations. This results in a higher order of convergence than the original method. One advantage is that the proposed method does not need any additional conditions, so it is efficient to implement. The next section will review the explicit error inhibiting block one-step method. In Sect. 3, we will explain the RBF interpolation. In Sect. 4, we show how the new method can be derived followed by Sect. 5 where numerical results are provided verifying that the convergence rate of the proposed method is increased by one order. A brief conclusion and an outline of our future research are presented in Sect. 6.

2 Error Inhibiting Block One-Step Method Consider the initial value problem for the first-order ODE below u (t) = f (t, u(t)), t  a

(1)

u(a) = ua

where we assume f (t, u) is uniformly Lipschitz continuous in u and continuous in t. We choose a value h for the step size and set tn = a + nh a discrete sequence in the time domain. Denote the numerical approximation of the solution u(tn ) by vn . Define the solution vector Un by "T ! Un = un+ s−1 , · · · , un+ 1 , un , s

s

where un+ j = u(tn+j h/s ) is the exact solution at t = tn + s The corresponding approximation vector Vn is defined as

jh s

for j = 0, · · · , s − 1.

"T ! Vn = vn+ s−1 , · · · , vn+ 1 , vn . s

s

In [3], the scheme is formulated as Vn+1 = QVn where the operator Q is represented by the following Q = A + hBf

(2)

An Adaptive Error Inhibiting Block One-Step Method for Ordinary Differential. . .

109

and A, B ∈ Rs×s . There are 4 sufficient conditions imposed on the matrices A and B in order to be error inhibiting: 1. rank(A) = 1. 2. The only non-zero eigenvalue of A is 1 and its corresponding eigenvector is [1, · · · , 1]T . 3. A can be diagonalized. 4. The matrices A and B are constructed such that when the local truncation error is multiplied by the discrete solution operator, we have ||Qτν ||  O(h) · ||τν ||. This is accomplished by requiring that the leading order term of the local truncation error is in the eigenspace of A associated with the zero eigenvalue. We derive those matrices of A and B with symbolic computation. As an example of the derivation of the error inhibiting method, we consider the construction of the scheme with s = 2. The solution vector is then Un = [un+1/2 , un ]T , and the corresponding approximation vector is given by Vn = [vn+1/2 , vn ]T . In order to satisfy those conditions listed above we first select 9 : 1−υ υ A= , 1−υ υ

(3)

which can be diagonalized as 9

: 9 :9 :9 : υ 1−υ υ υ −1 υ 1 0 −1 υ−1 A= = . 1−υ υ υ −1 υ −1 0 0 1 −1

(4)

Then conditions 1, 2 and 3 are satisfied. Further suppose that : b11 b12 . B= b21 b22 9

(5)

110

J. Gu and J.-H. Jung

Then Vn+1

: :9 : 9 9 1−υ υ b11 b12 fn+1/2 Vn + h = 1−υ υ b21 b22 fn

(6)

where fn+1/2 = f (tn+1/2 , vn+1/2 ) and fn = f (tn , vn ). The components of Vn+1 are vn+3/2 = (1 − υ)vn+1/2 + υvn + h(b11 fn+1/2 + b12fn ), vn+1 = (1 − υ)vn+1/2 + υvn + h(b21 fn+1/2 + b22fn ). We write each difference equation in the form of error normalized by the step size and then insert the exact solutions to the ODE into the difference equation. Expanding un+3/2 , un+1 and un+1/2 around t = tn in Taylor series gives the local truncation error τ n = (τn+1/2 , τn )T , where τn+1/2 =

1 1 (2 − 2b11 − 2b12 + υ)un + (8 − 4b11 + υ)un h 2 8 1 2 3 + (26 − 6b11 + υ)u(3) n h + O(h ), 48 (7)

τn =

1 1 (1 − 2b21 − 2b22 + υ)un + (3 − 4b21 + υ)un h 2 8 1 2 3 + (7 − 6b21 + υ)u(3) n h + O(h ). 48

(8)

Vanishing the coefficients of the constant term and the term h in (7) and (8), and υ , the equating the quotient of the coefficient of the terms h2 in (7) and (8) to υ−1 condition 4 is satisfied. Finally we have the desired scheme as in [3] Vn+1

: 9 : 9 :9 1 −1 7 h 55 −17 fn+1/2 = , Vn + fn 6 −1 7 24 25 1

(9)

and correspondingly the local truncation error is 2nd order convergent as expected 9 : 23 7 (3) 2 τn = u h + O(h3 ). 576 1 n

(10)

An Adaptive Error Inhibiting Block One-Step Method for Ordinary Differential. . .

111

3 RBF Interpolation Now we briefly explain the RBF interpolation in one dimension. Suppose that for a domain  ⊂ R, a data set {(xi , ui )}N i=0 is given where ui is the value of the unknown function u(x) at x = xi ∈ . We use the RBFs φ :  → R defined by φi (x) = φ(|x − xi |, #i ), where |x − xi | is the distance between x and xi and #i is a free parameter. The reconstruction of a function, u(x), is then made by a linear combination of RBFs INR u(x) =

N 

λi φ(|x − xi |, #i ),

(11)

i=0

where λi are the expansion coefficients to be determined. Using the interpolation condition INR u(xi ) = ui , i = 0, · · · , N, we could find the expansion coefficients λi by solving the linear system ⎡

⎤ ⎡ ⎤ ⎡ ⎤ φ(|x0 − x0 |, #0 ) φ(|x0 − x1 |, #1 ) · · · φ(|x0 − xN |, #N ) λ0 u0 ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ φ(|x1 − x0 |, #0 ) φ(|x1 − x1 |, #1 ) · · · φ(|x1 − xN |, #N ) ⎥ ⎢ λ1 ⎥ ⎢ u1 ⎥ ⎢ ⎥ · ⎢ . ⎥ = ⎢ . ⎥. .. .. .. ⎢ ⎥ ⎢ . ⎥ ⎢ . ⎥ . . . ⎣ ⎦ ⎣ . ⎦ ⎣ . ⎦ λN uN φ(|xN − x0 |, #0 ) φ(|xN − x1 |, #1 ) · · · φ(|xN − xN |, #N ) (12) If we choose the multiquadric RBF with all the free parameters equal, then the interpolation matrix, A, becomes a symmetric matrix with all diagonal entries 1, ⎤ , , 2 (x − x )2 · · · 1 + # 2 (x − x )2 1 1 + # 0 1 0 N , ⎥ ⎢, 1 · · · 1 + # 2 (x1 − xN )2 ⎥ ⎢ 1 + # 2 (x1 − x0 )2 ⎥. A=⎢ .. .. .. ⎥ ⎢ ⎦ ⎣, . . . , 1 + # 2 (xN − x0 )2 1 + # 2 (xN − x1 )2 · · · 1 (13) ⎡

Consider the case of three equally spaced nodes x0 , x1 , x2 with x0 < x1 < x2 . Let h be the grid spacing. Then the linear system becomes ⎡

⎤ ⎡ ⎤ ⎡ ⎤ √ √ 2 h2 2 h2 1 1 + # 1 + 4# u0 λ0 √ ⎢√ ⎥ ⎢ ⎥ ⎢ ⎥ 1 + # 2 h2 ⎦ · ⎣λ1 ⎦ = ⎣u1 ⎦ . ⎣√ 1 + # 2 h2 √ 1 λ2 u2 1 + 4# 2 h2 1 + # 2 h2 1

(14)

112

J. Gu and J.-H. Jung

By the closed-form expression for the RBF interpolant in [4], I2R u(x) =

2  i=0

ui det(Ai (x)). det(A)

(15)

where Ai (x), a 3 × 3 matrix, is obtained by replacing the ith row of A with the row vector " !, , , 1 + # 2 (x − x0 )2 1 + # 2 (x − x1 )2 1 + # 2 (x − x2 )2 . Differentiating the interpolant, we obtain the first-order derivative  ui d d R I2 u(x) = · det(Ai (x)). dx det(A) dx 2

(16)

i=0

We then estimate the derivative of u at x = x1 as we do in polynomial interpolation for the central difference formula: √ 1 + 4# 2 h2 + 1 d R I2 u(x1 ) = (u2 − u0 ). (17) √ dx 4h 1 + # 2 h2 By employing the Taylor expansion of the quotient on the right-hand side of (17), we have  d R h 1 I u(x1 ) = + # 2 + O(h3 ) (u2 − u0 ). (18) dx 2 2h 4 The main feature of the RBF method is that it contains a free parameter, #, which we could make use of to further inhibit the errors. In the following section, we will show that using the parameter # coupled with hp terms, where p  2, we can increase the order of local truncation error and further promote the order of global error by adopting the error inhibiting scheme.

4 Construction of the Adaptive Error Inhibiting Scheme Following the main feature of the RBF method explained in the preceding section, we try to establish a similar explicit block one-step scheme that provides a higher order of convergence by adding one more block of the free parameters #1 and #2

An Adaptive Error Inhibiting Block One-Step Method for Ordinary Differential. . .

113

coupled with hp term. With p = 3, we have :9 : 9 :9 : : 9 fn+1/2 1−υ υ b11 b12 fn+1/2 3 0 #1 +h . = Vn + h b21 b22 fn 0 #2 fn 1−υ υ 9

Vn+1

(19)

We measure the one-step error normalized by the step size as in Sect. 2. Expanding un+3/2 , un+1 and un+1/2 around t = tn in Taylor series again yields the local truncation error τ n = [τn+1/2 , τn ]T , where τn+1/2 =

1 1 (2 − 2b11 − 2b12 + υ)un + (8 − 4b11 + υ)un h + 2 8

 1 1 (3) (4) (80 − 8b11 + υ)un h3 + O(h4 ), −#1 un + (26 − 6b11 + υ)un h2 + 48 384

(20) τn =

1 1 (1 − 2b21 − 2b22 + υ)un + (3 − 4b21 + υ)un h + 2 8

 1 1 (3) (4) (15 − 8b21 + υ)un h3 + O(h4 ). −#2 un + (7 − 6b21 + υ)un h2 + 48 384

(21) Annihilating the first two terms in (20) and (21), and equating the quotient of the υ coefficient of the terms h3 in (20) and (21) to υ−1 , we have the scheme Vn+1

: 9 :9 : 9 : 9 :9 1 −1 8 h 64 −20 fn+1/2 0 # f 1 n+1/2 + h3 . Vn + = fn 0 #2 fn 7 −1 8 28 29 1 (22)

We can easily check that the scheme (22) satisfies those four conditions in Sect. 2. Further annihilating the coefficients of the term h2 , we get the optimal values of #1 and #2 : (3)

#1 =

47un , 168un

(23)

#2 =

9u(3) n . 224un

(24)

114

J. Gu and J.-H. Jung

Our new scheme has the truncation error 9 : 55 8 (4) 3 u h + O(h4 ). τn = 2688 1 n

(25)

(3)

Note that in our new scheme, we need the value of un at each step. This higher order derivative can be computed by repeated differentiation of the function f on the right-hand side of (1) twice. However, we choose to estimate the third-order derivative. For un , we use the given condition from (1), i.e. u (t) = f (t, u(t)). For the third-order derivative u(3) n , we employ the second-order central difference formula for f  (t, u(t)) at t = tn as  u(3) n = f (tn , un ) ≈

4(fn+1/2 + 2fn − fn−1/2 ) , h2

(26)

where fn+1/2 , fn and fn−1/2 are given values. For this computation, no additional conditions are necessary. The truncation error is still third order accurate, O(h3 ), as in (25), so by the error inhibiting strategy we end up with a global error that is O(h4 ), which will soon be confirmed in the following section. We conclude this section with a comparison of three methods. For DIMSIM of type 3, 9 : : : 9 :9 9 :9 1 7 −3 vn+1 h 9 −7 fn+1 vn+2 = + , vn+1 vn fn 4 7 −3 8 −3 −3 two steps vn and vn+1 are employed to update the step vn+1 and obtain the step vn+2 . For error inhibiting scheme, 9

: : : 9 :9 9 :9 1 −1 7 vn+1/2 h 55 −17 fn+1/2 vn+3/2 = + , vn+1 vn fn 6 −1 7 24 25 1

two steps vn and vn+1/2 are involved to generate the next two steps vn+1 and vn+3/2 . For our method (if we utilize (26) and substitute (23), (24) for respective #1 and #2 in (22) to avoid the zero denominator), ⎡

⎤ ⎡ vn+3/2 −1 8 ⎢ ⎥ 1⎢ ⎣ vn+1 ⎦ = ⎣−1 8 7 vn+1/2 1 0

⎤ ⎤ ⎡ ⎤⎡ ⎤⎡ 0 572 −496 188 vn+1/2 fn+1/2 h ⎢ ⎥ ⎥ ⎥⎢ ⎥⎢ 0⎦ ⎣ vn ⎦ + ⎣201 −48 27 ⎦ ⎣ fn ⎦ , 168 vn−1/2 fn−1/2 0 0 0 0

we use previous three steps vn−1/2 , vn and vn+1/2 to evolve the next two steps vn+1 and vn+3/2 . In [5] the stability analysis has been done for the adaptive radial basis function methods for IVPs and it has been shown that some adaptive methods have a

An Adaptive Error Inhibiting Block One-Step Method for Ordinary Differential. . .

115

better stability condition than the original ones. However, it seems that the adaptive error inhibiting method is more computationally expensive than the original one when the approximation of (26) is used.

5 Numerical Results We start with the nonlinear first-order differential equation used in [3] u = −u2 , t  0

(27)

u(0) = 1.

The exact solution of the example is u(t) = 1/(t +1). The left figure of Fig. 1 shows the global errors at the time t = 1 versus N, the number of steps, in logarithmic scale for the type-3 DIMSIM (blue), the original error inhibiting scheme (red) and our proposed method (green). As seen in the figure, our proposed method is the most accurate among those three methods and yields high order convergence which is 4th order. Table 1 shows the convergence with N for (27). The type-3 DIMSIM yields the 2nd order accuracy, the original error inhibiting scheme yields the 3rd order accuracy and our proposed method yields the 4th order accuracy. Next we consider the following problem used in [6] where the solution changes rapidly between [−2, 2] u = −4t 3 u2 , t  −10

(28)

u(−10) = 1/10001. 10 -2 10 0 10 -4

Global Error

Global Error

10 -2 10 -6

10 -8

10 -4

10 -6

DIMSIM3 EIS 3

10 -10

DIMSIM3 EIS 3

EIS with h Slope = -2 Slope = -3 Slope = -4

10 -8

10 1

10 2

N

EIS with h Slope = -2 Slope = -3 Slope = -4

10 3

N

Fig. 1 Global error versus N in logarithmic scale. Left: (27). Right: (28). Blue: DIMSIM (DIMSIM3) 2nd order. Red: error inhibiting scheme (EIS) 3rd order. Green: our proposed method (EIS with h3 ) 4th order

116

J. Gu and J.-H. Jung

Table 1 Global error and order of convergence for u = −u2 with u(0) = 1 Method DIMSIM type-3

Error inhibiting scheme

Error inhibiting scheme with h3 term

N 10 20 40 80 160 320 10 20 40 80 160 320 10 20 40 80 160 320

Global error 6.60E−3 1.60E−3 3.82E−4 9.41E−5 2.34E−5 5.82E−6 2.17E−4 2.89E−5 3.73E−6 4.74E−7 5.97E−8 7.50E−9 2.71E−5 2.24E−6 1.64E−7 1.11E−8 7.22E−10 4.61E−11

order 2.0702 2.0402 2.0208 2.0105 2.0053 2.9118 2.9536 2.9763 2.9880 2.9940 3.5935 3.7781 3.8833 3.9400 3.9698

The exact solution is u(t) = 1/(t 4 + 1). The right figure of Fig. 1 shows the global errors at t = 0 versus N in logarithmic scale for the type-3 DIMSIM (blue), the original error inhibiting method (red) and our proposed method (green). We verify again that our proposed method is indeed the most accurate and yields the highest order of convergence. Table 2 shows the convergence with N for (28). Although the type-3 DIMSIM does not reveal the 2nd order accuracy in the beginning, it eventually exhibits the order of accuracy as expected. The original error inhibiting scheme is 3rd order accurate and our proposed method 4th order accurate.

6 Conclusions In this note, we modified and improved the original error inhibiting block onestep method proposed in [3] by introducing a free parameter. By exploiting the parameter, the local truncation error is further reduced resulting in higher order of the global error. It is numerically demonstrated that, with the proposed method, the local truncation error is of the 3rd order and the global error of the 4th order. As mentioned in Sect. 4, we will investigate the stability of the error inhibiting method and our proposed method as well as relaxing the fourth constraint in error inhibiting method in our future research.

An Adaptive Error Inhibiting Block One-Step Method for Ordinary Differential. . .

117

Table 2 Global error and order of convergence for u = −4t 3 u2 with u(−10) = 1/10001 Method DIMSIM type-3

Error inhibiting scheme

Error inhibiting scheme with h3 term

N 200 400 800 1600 3200 6400 200 400 800 1600 3200 6400 200 400 800 1600 3200 6400

Global error 9.05E−1 7.24E−1 4.07E−1 1.49E−1 4.24E−2 1.10E−2 1.86E−1 2.80E−2 3.60E−3 4.50E−4 5.63E−5 7.03E−6 1.14E−2 6.57E−4 3.94E−5 2.41E−6 1.49E−7 9.91E−9

Order 0.3221 0.8293 1.4476 1.8158 1.9475 2.7294 2.9639 2.9965 3.0002 3.0005 4.1132 4.0620 4.0307 4.0123 3.9122

Acknowledgements The authors thank Adi Ditkowski for introducing the error inhibiting method to us and communicating with us on the subject. The research is partially supported by Ajou University.

References 1. Butcher, J.C.: Diagonally-implicit multi-stage integration method. Appl. Numer. Math. 11, 347– 363 (1993) 2. Ditkowski, A.: High order finite difference schemes for the heat equation whose convergence rates are higher than their truncation errors. In: R.M. Kirby et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2014, pp. 167–178. Springer, Switzerland (2015) 3. Ditkowski, A., Gottlieb, S.: Error inhibiting block one-step schemes for ordinary differential equations. J. Sci. Comput. 73, 691–711 (2017) 4. Fornberg, B., Weight, G., Larsson, E.: Some observations regarding interpolants in the limit of flat radial basis functions. Comput. Math. Appl. 47, 37–55 (2004) 5. Gu, J. Jung, J.-H.: Adaptive radial basis function methods for initial value problems. J. Sci. Comput. 82, 47 (2020). https://doi.org/10.1007/s10915-020-01140-0 6. Sauer, T.: Numerical Analysis, 2nd edn. Pearson, New York (2012) 7. Shampine, L.F., Watts, H.A.: Block implicit one-step methods. Math. Comp. 23, 731–740 (1969)

118

J. Gu and J.-H. Jung

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Hermite Methods in Time Rujie Gu and Thomas Hagstrom

1 Introduction Over the past decade a number of works have appeared which exploit the unique properties of Hermite-Birkhoff interpolation in space to construct arbitraryorder discretization methods for hyperbolic [1, 2, 4–7, 10, 14, 16–19] as well as Schrödinger [3] equations. The precise form of the interpolant in a single cell, which here we write in one dimension labelled t, is u(t) ≈ Iu(t) ∈ Π 2m+1 ,

t ∈ (tj −1 , tj ),

dk dku Iu(t ) = k (t ); k = 0, . . . , m, k dt dt

 = j − 1, j,

(1) (2)

where Π 2m+1 denotes the polynomials of degree 2m + 1. (In higher dimensions one uses a tensor-product cell interpolant based on vertex data consisting of mixed derivatives of order through m in each Cartesian coordinate.) In contrast, there has been little work on analogous methods for time discretization. A recent exception is the manuscript by Liu et al. [15]. They develop methods for second-order semilinear hyperbolic equations using interpolants of the form (1)–(2) combined with a reformulation of the evolution problem using exact solutions of the linear part. They demonstrate excellent long-time performance. The outline of the paper is as follows. In Sect. 2 we list a few properties of piecewise Hermite-Birkhoff interpolation. In Sect. 3 we construct the time-stepping R. Gu · T. Hagstrom () Southern Methodist University, Dallas, TX, USA e-mail: [email protected]; [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_8

119

120

R. Gu and T. Hagstrom

schemes and establish some basic results, with a few numerical experiments described in Sect. 4.

2 Basic Properties of Hermite-Birkhoff Interpolation Hermite interpolants have a number of interesting properties which make them very attractive for the solution of differential equations; see, e.g., [2]. Here we will mainly use the simplest. Precisely, for t ∈ (tj −1 , tj ), the Peano representation of the local error can be easily derived by noting that e = u − Iu solves the two point boundary value problem d 2m+2 e d 2m+2 u d k e = , = 0, t = tj −1 , tj , k = 0, . . . m. dt 2m+2 dt 2m+2 dt k

(3)

Thus  e(t) =

tj

Kj (t, s) tj−1

d 2m+2 u (s)ds, dt 2m+2

(4)

where the kernel Kj is the Green’s function for (3). Simple scaling arguments combined with the transformation t = tj −1 + zhj then show that e = O(h2m+2 ) j where hj = tj − tj −1 is the time step. A fundamental feature of piecewise Hermite interpolation is the following orthogonality property. For any functions v(t), w(t) 

tj tj−1

  d m+1 w − Iw d m+1 Iv (t) · (t)dt = 0, dt m+1 dt m+1

(5)

which in particular implies that interpolation reduces the H m+1 seminorm.

3 Time-Stepping Methods We begin by considering the initial value problem for a first-order system ordinary differential equations: du = f (u, t), dt

u(t0 ) = u0 ,

u(t) ∈ Rd .

(6)

Given a discrete time sequence tj > tj −1 , j = 1, . . . , N, with time steps hj = tj − tj −1 we write down the Picard integral formulation of the time evolution over

Hermite Methods in Time

121

a single step  u(tj ) = u(tj −1 ) +

tj

f (u(s), s)ds.

(7)

tj−1

The construction of our time integration formula proceeds in three steps. We denote by vj the approximation to u(tj ). 1. Given vj −1 and assuming for the moment that vj is known, use the differential equation to compute m scaled derivatives of its solution, V (t), satisfying V (t ) = v ,  = j, j − 1. Setting F (t) = f (V (t), t),

(8)

these are recursively defined by the formula d k−1 F d k V (t ) = (t ), k dt dt k−1

k = 0, . . . m.

(9)

2. Construct the Hermite-Birkhoff interpolant of this data; that is the polynomial Pj −1/2 (t; vj −1 , vj ) of degree 2m + 1, satisfying d k Pj −1/2 d k V (t ; v , v ) = (t );  = j − 1, j, k = 0, . . . , m..  j −1 j dt k dt k

(10)

3. Approximate (7) by replacing u(t ) by v and replacing the integral by a q + 1point quadrature rule with f evaluated at the Hermite interpolant: vj = vj −1 + hj

q 

wk f (Pj −1/2 (tj,k ; vj −1 , vj )).

(11)

k=0

4. Solve (11) for vj . Note that this is a system of d nonlinear equations for any m; that is, unlike standard implicit Runge–Kutta methods, the size of the nonlinear system is independent of the order. We remark that we have not studied in detail the unique solvability of (11) in the stiff case. In our numerical experiments we used the solution at the current time step as an initial approximation for Newton iterations and simply accepted the solution to which the iterates converged. To emphasize the ideas we write down some specific examples of methods with m = 1 and m = 2 making the simplifying assumption of autonomy; that is f = f (u). The derivation of methods of arbitrary order is straightforward and the formulas can be trivially obtained using software capable of symbolic computations. To apply them at higher order one must evaluate higher derivatives of f , which is also possible using automatic differentiation tools [11].

122

R. Gu and T. Hagstrom

Example (m = 1) Set τ=

t − tj −1 . hj

Now the interpolant Pj −1/2 (t; vj −1 , vj ) is given by: Pj −1/2 (t; vj −1 , vj ) =

3 

ak τ k ,

(12)

k=0

where a0 = vj −1 , a1 = hj f (vj −1 ),     a2 = 3 vj − vj −1 − hj 2f (vj −1 ) + f (vj ) ,

(13)

    a3 = −2 vj − vj −1 + hj f (vj −1 ) + f (vj ) . We next introduce a quadrature rule which is exact for polynomials of degree 3. Possible choices include the 2-point Gauss-Legendre (14) rules, or the 3-point Gauss-Radau (15) or Gauss-Lobatto rules. Note that by using two different rules we obtain a possible error indicator. Here are the two different methods used below. Note that the methods are identical if f is linear. vj = vj −1 +

   hj   f Pj −1/2 (α− ; vj −1 , vj ) + f Pj −1/2 (α+ ; vj −1 , vj ) , 2 (14)

    hj  β+ f Pj −1/2 (γ− ; vj −1 , vj ) + β− f Pj −1/2 (γ+ ; vj −1 , vj ) 36   +4f vj , (15)

vj = vj −1 +

#

$ √ √ 1 4± 6 , β± = 16 ± 6. α± = 1 ± √ , γ± = 10 3

(16)

A time step is executed by solving the nonlinear system, (14) or (15), for vj . Example (m = 2) Now we also need the second time derivative of u, d d 2u = f (u) = J (u)f (u), 2 dt dt

(17)

Hermite Methods in Time

123

where J (u) is the Jacobian derivative. The Hermite interpolant can now be written: Pj −1/2 (t; vj −1 , vj ) =

5 

ak τ k ,

(18)

k=0

where a0 = vj −1 , a1 = hj f (vj −1 ), a2 =

h2j 2

J (vj −1 )f (vj −1 ),

(19)

    h2   j −3J (vj −1 )f (vj −1 )+J (vj )f (vj ) , a3 =10 vj −vj −1 −hj 6f (vj −1 )+4f (vj ) + 2     h2   j 3J (vj −1 )f (vj −1 )−J (vj )f (vj ) , a4 =−15 vj −vj −1 +hj 8f (vj −1 )+7f (vj ) + 2     h2   j −J (vj −1 )f (vj −1 )+J (vj )f (vj ) . a5 = 6 vj −vj −1 −3hj f (vj −1 )+f (vj ) + 2

Again we can now use, for example, the 3-point Gauss-Legendre or 4-point Gauss-Radau quadrature rules to produce the equation we must solve for vj .

3.1 Stability and Consistency The consistency of the method is a straightforward consequence of its construction, and its linear stability properties can also be established. Theorem 1 Assume that the quadrature rule has positive weights and is exact for polynomials of degree 2m + 1. Then the implicit Hermite method is A-stable and accurate of order 2m + 2. Proof Assume that f is smooth and that u(t) ∈ C 2m+2 (0, T ). Using (4), standard estimates for quadrature errors, and the Picard formula (7) we find for the truncation error u(tj ) − u(tj −1 )  − wk f (Pj −1/2 (tj,k ; u(tj −1 ), u(tj ))), (20) hj k ⏐ ⏐ ⏐ t ⏐  j ⏐ ⏐ 1 ⏐ ⏐ |τj | = f (u(s))ds − h w f (P (t ; u(t ), u(t ))) j k j −1/2 j,k j −1 j ⏐ ⏐ hj ⏐ tj−1 ⏐ τj =

k

124

R. Gu and T. Hagstrom

⏐ ⏐ ⏐ t j ⏐  ⏐ ⏐ ⏐ ⏐ f (u(s))ds − h w f (u(t )) j k j,k ⏐ ⏐ ⏐ tj−1 ⏐ k ⏐ ⏐ ⏐ ⏐ ⏐  ⏐ ⏐ +⏐ wk f (u(tj,k )) − f (Pj −1/2 (tj,k ; u(tj −1 ), u(tj ))) ⏐ ⏐ ⏐ k ⏐

1 ≤ hj

. ≤ Ch2m+2 j

(21)

Now consider the Dahlquist test problem, f (u) = λu. In this case all quadrature rules which are exact for the Hermite interpolant produce the same method. As interpolation is linear, we have that the coefficients of the interpolant are linear combinations hkj λk vj −1 and hkj λk vj , k = 0, . . . m. The Picard integral then increases the powers of hj λ by one so that the implicit system (11) can be rearranged to: Q+ (hj λ)vj = Q− (hj λ)vj −1 ⇒ vj =

Q− (hj λ) vj −1 , Q+ (hj λ)

(22)

where Q± (hj λ) are polynomials of degree m + 1. Consistency implies ehj λ =

  Q− (hj λ) + O (hj λ)2m+3 . Q+ (hj λ)

(23)

The only rational function of the given degree with this accuracy is the diagonal Padé approximant. We thus conclude that our methods are A-stable [12].#

4 Numerical Experiments Our first experiments treat standard problems from the ode literature and are restricted to the fourth and sixth order methods described above with either GaussLegendre or Gauss-Radau quadrature. Our practical implementations employ the classical Aitken algorithm adapted to Hermite interpolation to directly evaluate Pj −1/2 (tj,k , vj −1 , vj ) and solve (11) using Newton’s method with the Jacobian of the implicit system approximated by finite differences. For adaptive computations we 1. Compute vj using the Gauss-Radau-based formulas, 2. Compute a residual, ρj , by substituting vj into the Gauss-Legendre-based formulas.

Hermite Methods in Time

125

We then adjust the time step by the simple rule # hj +1 =

tol ρj

$1/(2m+3) (24)

hj ,

while also imposing a minimum time step. Our final experiment examines the use of the method for evolving spectral discretizations of initial-boundary value problems for the Schrödinger equation.

4.1 Arentsorf Orbit We first consider the problem of computing a periodic solutions of the restricted three-body problem which we reformulate as a first-order system of four variables: d 2 y1 dy2 = y1 + 2 − (1 − μ)  2 dt dt d 2 y2 dy1 = y2 − 2 − (1 − μ)  2 dt dt

y1 + μ (y1 + μ)2 + y22 y2 (y1 + μ)2 + y22

3/2 − μ 

3/2 − μ 

μ = 0.012277471, y1 (0) = .994,

y1 − (1 − μ) (y1 − (1 − μ))2 + y22 y2 (y1 − (1 − μ))2 + y22

3/2 ,

3/2 ,

dy1 (0) = y2 (0) = 0, dt

dy2 (0) = −2.01585106379082 . . ., T = 17.06521656015796 . . .. dt (For graphs of the solution see [13, Ch. II].) We note that this problem is not considered to be stiff. The main difficulty is a need for very small time steps when the orbits approach the singularities of f . However, we use it to verify convergence at the design order when (woefully inefficient) uniform time steps are employed and to test the utility of our naive time step adaptivity algorithm. Results for fixed (small) time steps are displayed in Table 1. We observe that convergence is at design order and that the results for the two quadrature formulas are comparable, though the fourth order Radau method is somewhat more accurate than Gauss-Legendre with roles reversed at sixth order. The sixth order methods are more accurate with larger time steps. The error is simply , (y1 (T ) − y1 (0))2 + (y2 (T ) − y2 (0))2 . Results for adaptive computations with m = 2 are shown in Table 2. Obviously, the adaptive methods lead to a very significant reduction in the number of time steps; an accuracy of 10−7 is achieved with 264 steps of the adaptive method

Gauss-Legendre m = 1 h Error 5.69(−4) 2.30(−4) 4.88(−4) 1.24(−4) 4.27(−4) 7.26(−5) 3.79(−4) 4.53(−5) 3.41(−4) 2.97(−5) 3.10(−4) 2.03(−5) 2.84(−4) 1.43(−5)

4.01 4.01 4.01 4.01 4.00 4.00

Rate

Gauss-Radau m = 1 h Error 5.69(−4) 9.65(−5) 4.88(−4) 5.21(−5) 4.27(−4) 3.05(−5) 3.79(−4) 1.91(−5) 3.41(−4) 1.25(−5) 3.10(−4) 8.54(−6) 2.84(−4) 6.03(−6) 4.00 4.00 4.00 4.00 4.00 4.00

Rate

Table 1 Convergence with fixed time steps for the Arentsorf orbit problem Gauss-Legendre m = 2 h Error 8.53(−4) 2.98(−6) 6.83(−4) 7.82(−7) 5.69(−4) 2.62(−7) 4.88(−4) 1.04(−7) 4.27(−4) 4.66(−8) 3.79(−4) 2.30(−8) 3.41(−4) 1.22(−8) 6.00 6.00 6.00 6.00 6.00 6.00

Rate

Gauss-Radau m = 2 h Error 8.53(−4) 4.46(−6) 6.83(−4) 1.16(−6) 5.69(−4) 3.88(−7) 4.88(−4) 1.53(−7) 4.27(−4) 6.87(−8) 3.79(−4) 3.39(−8) 3.41(−4) 1.80(−8)

6.03 6.02 6.01 6.01 6.01 6.00

Rate

126 R. Gu and T. Hagstrom

Hermite Methods in Time

127

Table 2 Time steps and error as a function of tolerance for adaptive solutions of the Arentsorf problem with m = 2 van der Pol y

2.5

1

Tol 10−6 10−8 10−10 - tol=1.0e-08

Steps 65 136 264

time steps - tol=1.0e-08

10-2

2

Error 8.21(−3) 1.85(−5) 1.15(−8)

10-3

1.5

10-4

1 0.5

10-5

0 -0.5

10-6

-1

10-7

-1.5 10-8

-2 -2.5

0

2

4

6

8

10

12

10-9

0

2

4

6

8

10

12

Fig. 1 Solution and time step history for the van der Pol oscillator with tolerance 10−8

while 35,000 uniform steps are required. Due to the sensitivity of the problem, the global error is much larger than the error tolerance, but is reduced in proportion to it.

4.2 Van der Pol Oscillator Our second example is the van der Pol oscillator problem, which again we rewrite as a first order system:

 d 2y −1 2 dy −y , (1 − y ) =# dt dt 2 # = 10−6 , y(0) = 2,

(25)

dy (0) = 0. dt

We solve up to T = 11 using the adaptive method with m = 2. We plot the solution and the time step histories for a tolerance of 10−10 in Fig. 1. Note that very small steps are needed to resolve the fast transitions, while the problem is quite stiff in the regions where y is nearly constant. Plots for the other tolerances tested, 10−6 and 10−10, are similar though the number of time steps required varies.

128

R. Gu and T. Hagstrom

10-2

Nonlinear Schrodinger: m=5

Nonlinear Schrodinger: h=.01

100 10-2

10-4 10

Relative Error

Relative Error

10-4 -6

10-8 10-10 10-12 0

h=.03 h=.02 h=.015 h=.01 h=.0075 h=.006 h=.005

0.5

10-6 10-8 10-10

m=1 m=2 m=3 m=4 m=5

10-12 1

1.5 t

2

2.5

3

10-14

0

0.5

1

1.5 t

2

2.5

3

Fig. 2 Left: Relative errors for NLS with various time steps and m = 3. Right: Relative errors for NLS with h = .01 and varying m

4.3 Schrödinger Equation Lastly, we apply the method to evolve a Fourier pseudospectral discretization of the nonlinear Schrödinger equation. Precisely we consider the real problem  ∂ 2w  ∂v = − 2 − v 2 + w2 w, ∂t ∂x

 ∂ 2v  ∂w = 2 + v 2 + w2 v, ∂t ∂x

(26)

∂v for x ∈ (−8, 8), t ∈ (0, 3) with periodic boundary conditions ∂x = ∂w ∂x = 0 at x = ±1. We approximate the periodization of the exact solitary wave solution

√ 50 cos (rx − st) · sech(5(x − ct)), √ w(x, t) = 50 sin (rx − st) · sech(5(x − ct)), v(x, t) =

(27)

with c = 2π, r = π , s = π 2 − 25. We note that the amplitude of the solitary wave is reduced by about 17 digits at a distance of 8 from its peak so that the interaction with periodic copies is negligible over the simulation time. We use 512 Fourier modes in the computation of the derivatives and experiments show that this is sufficient to represent the solitary wave to machine precision. The implicit system was solved using Newton iterations each time step. In Fig. 2 we present results for m = 3 (8th order) with varying time step and for m varying from 1 to 5 (order 4 through 12) with h = 10−2 . In both cases we observe rapid convergence. We also tabulate the errors at the final time and calculate the convergence rates when m = 3 in Table 3. The results are clearly consistent with the design order.

Hermite Methods in Time Table 3 Relative errors for the Fourier pseudospectral discretization of the NLS (26) with solitary wave solution (27)

129 m=3 h 3.0(−2) 2.0(−2) 1.5(−2) 1.0(−2) 7.5(−3) 6.0(−3) 5.0(−3)

Error 4.0(−3) 6.7(−5) 4.9(−6) 1.8(−7) 2.2(−8) 3.2(−9) 9.4(−10)

Rate 10.1 9.1 8.1 7.3 8.7 6.8

h = 0.01 m Error 1 1.2(−2) 2 5.0(−5) 3 1.8(−7) 4 8.6(−10) 5 1.2(−10)

5 Conclusions and Future Work In conclusion, we have demonstrated that Hermite-Birkhoff interpolation can be used to develop singly-implicit A-stable timestepping methods of arbitrary order. A number of possible generalizations and improvements to the method are possible. These include 1. Stability analysis for variable coefficient or nonlinear problems using the projection properties (5); 2. Improved time step/order adaptivity; 3. Preconditioning of the implicit system for applications to partial differential equations such as spectral/pseudospectral discretizations of equations of Schrödinger type (e.g. integration preconditioners [8, 9]); 4. Development of IMEX schemes combining Hermite and Taylor polynomials. Acknowledgements This work was supported in part by NSF Grant DMS-1418871. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the National Science Foundation.

References 1. Appelö, D., Hagstrom, T.: On advection by Hermite methods. Pac. J. Appl. Math. 4,125–139 (2012) 2. Appelö, D., Hagstrom, T.: Solving PDEs with Hermite Interpolation. Lecture Notes in Computational Science, pp. 31–49. Springer, Berlin (2015) 3. Appelö, D., Kreiss, G., Wang, S.: An explicit Hermite-Taylor method for the Schrödinger equation. Commun. Comput. Phys 21, 1207–1230 (2017) 4. Appelö, D., Hagstrom, T., Vargas, A.: Hermite methods for the scalar wave equation. SIAM J. Sci. Comput. 40, A3902–A3927 (2018) 5. Chen, R., Hagstrom, T.: P-adaptive Hermite methods for initial value problems. ESAIM: Math. Model. Numer. Anal. 46, 545–557 (2012) 6. Chen, X., Appelö, D., Hagstrom, T.: A hybrid Hermite-discontinuous Galerkin method for hyperbolic systems with application to Maxwell’s equations. J. Comput. Phys. 257, 501–520 (2014)

130

R. Gu and T. Hagstrom

7. Chidwagyai, P., Nave, J.-C., Rosales, R., Seibold, B.: A comparative study of the efficiency of jet schemes. Int. J. Numer. Anal. Model.-B 3, 297–306 (2012) 8. Coutsias, E., Hagstrom, T., Hesthaven, J., Torres, D.: Integration preconditioners for differential operators in spectral τ methods. Houst. J. Math. 1995, 21–38 (1996). Special issue: ICOSAHOM 9. Coutsias, E., Hagstrom, T., Torres, D.: An efficient spectral method for ordinary differential equations with rational function coefficients. Math. Comp. 65, 611–635 (1996) 10. Goodrich, J., Hagstrom, T., Lorenz, J.: Hermite methods for hyperbolic initial-boundary value problems. Math. Comp. 75, 595–630 (2006) 11. Griewank, A.: Evaluating Derivatives: Principles and Techniques of Algorithmic Differentiation. SIAM, Philadelphia (2000) 12. Hairer, E., Wanner, G.: Solving Ordinary Differential Equations II, Stiff and DifferentialAlgebraic Problems. Springer, New York (1996) 13. Hairer, E., Norsett, S., Wanner, G.: Solving Ordinary Differential Equations I, Nonstiff Problems. Springer, New York (1992) 14. Kornelus, A., Appelö, D.: Flux-conservative Hermite methods for simulation of nonlinear conservation laws. J. Sci. Comput. 76, 24–47 (2018) 15. Liu, C., Iserles, A., Wu, W.: Symmetric and arbitrarily high-order Birkhoff-Hermite time integrators and their long-time behaviour for solving nonlinear Klein-Gordon equations. J. Comput. Phys. 356, 1–30 (2018) 16. Seibold, B., Rosales, R., Nave, J.-C.: Jet schemes for advection problems. Discrete Contin. Dyn. Syst. Ser. B 17, 1229–1259 (2012) 17. Vargas, A., Chan, J., Hagstrom, T., Warburton, T.: GPU Acceleration of Hermite Methods for Simulation of Wave Propagation. Lecture Notes in Computational Science, pp. 357–368. Springer, Berlin (2017) 18. Vargas, A., Chan, J., Hagstrom, T., Warburton, T.: Variations on Hermite methods for wave propagation. Commun. Comput. Phys. 22, 303–337 (2017) 19. Vargas, A., Hagstrom, T., Chan, J., Warburton, T.: Leapfrog time-stepping for Hermite methods. J. Sci. Comput. 80, 289–314 (2019)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

HPS Accelerated Spectral Solvers for Time Dependent Problems: Part II, Numerical Experiments Tracy Babb, Per-Gunnar Martinsson, and Daniel Appelö

1 Introduction In this chapter describes a highly computationally efficient solver for equations of the form κ

∂u = Lu(x, t) + h(u, x, t), x ∈ Ω, t > 0, ∂t

(1)

with initial data u(x, 0) = u0 (x). Here L is an elliptic operator acting on a fixed domain Ω and h is lower order, possibly nonlinear terms. We take κ to be real or imaginary, allowing for parabolic and Schrödinger type equations. We desire the benefits that can be gained from an implicit solver, such as L-stability and stiff accuracy, which means that the computational bottleneck will be the solution of a sequence of elliptic equations set on Ω. In situations where the elliptic equation to be solved is the same in each time-step, it is highly advantageous to use a direct (as opposed to iterative) solver. In a direct solver, an approximate solution operator to the elliptic equation is built once. The cost to build it is typically higher than the cost required for a single elliptic solve using an iterative method such as multigrid, but the upside is that after it has been built, each subsequent solve is very fast. In this chapter, we argue that a particularly efficient direct solver to use in this context is a method obtained by combining a multidomain spectral collocation discretization (a

T. Babb · D. Appelö () University of Colorado, Boulder, CO, USA e-mail: [email protected]; [email protected] P.-G. Martinsson University of Texas, Austin, TX, USA e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_9

131

132

T. Babb et al.

so-called “patching method”, see e.g. Ch. 5.13 in [3]) with a nested dissection type solver. It has recently been demonstrated [1, 7, 12] that this combined scheme, which we refer to as a “Hierarchial Poincaré–Steklov (HPS)” solver, can be used with very high local discretization orders (up to p = 20 or higher) without jeopardizing either speed or stability, as compared to lower order methods. In this chapter, we investigate the stability and accuracy that is obtained when combining high-order time-stepping schemes with the HPS method for solving elliptic equations. We restrict attention to relatively simple geometries (mostly rectangles). The method can without substantial difficulty be generalized to domains that can naturally be expressed as a union of rectangles, possibly mapped via curvilinear smooth parameter maps. A longer version of this chapter with additional details is available at [2]. Also note that the conclusions are deferred to Part II of this paper (same issue).

2 The Hierarchical Poincaré–Steklov Method In this section, we describe a computationally efficient and highly accurate technique for solving an elliptic PDE of the form [Au](x) = g(x),

x ∈ Ω,

u(x) = f (x),

x ∈ Γ,

(2)

where Ω is a domain with boundary Γ , and where A is a variable coefficient elliptic differential operator [Au](x) = −c11 (x)[∂12 u](x) − 2c12 (x)[∂1 ∂2 u](x) − c22 (x)[∂22 u](x) + c1 (x)[∂1 u](x) + c2 (x)[∂2 u](x) + c(x) u(x) with smooth coefficients. In the present context, (2) represents an elliptic solve that is required in an implicit time-descretization technique of a parabolic PDE, as discussed in Sect. 1. For simplicity, let us temporarily suppose that the domain Ω is rectangular; the extension to more general domains is discussed in Remark 1. Our ambition here is merely to provide a high level description of the method; for implementation details, we refer to [1, 2, 7–9, 12, 13].

2.1 Discretization We split the domain Ω into n1 × n2 boxes, each of size h × h. Then on each box, we place a p × p tensor product grid of Chebyshev nodes, as shown in Fig. 1. We use collocation to discretize the PDE (2). With {x i }N i=1 denoting the collocation points,

HPS Accelerated Solvers: Part I, Numerical Experiments

133

Fig. 1 The domain Ω is split into n1 × n2 squares, each of size h × h. In the figure, n1 = 3 and n2 = 2. Then on each box, a p × p tensor product grid of Chebyshev nodes is placed, shown for p = 7. At red nodes, the PDE (2) is enforced via collocation of the spectral differentiation matrix. At the blue nodes, we enforce continuity of the normal fluxes. Observe that the corner nodes (gray) are excluded from consideration

the vector u that represents our approximation to the solution u of (2) is given simply by u(i) ≈ u(x i ). We then discretize (2) as follows: 1. For each collocation node that is internal to a box (red nodes in Fig. 1), we enforce (2) by directly collocating the spectral differential operator supported on the box, as described in, e.g., Trefethen [15]. 2. For each collocation node on an edge between two boxes (blue nodes in Fig. 1), we enforce that the normal fluxes across the edge be continuous. For instance, for a node x i on a vertical line, we enforce that ∂u/∂x1 is continuous across the edge by equating the values for ∂u/∂x1 obtained by spectral differentiation of the boxes to the left and to the right of the edge. For an edge node that lies on the external boundary Γ , simply evaluate the normal derivative at the node, as obtained by spectral differentiation in the box that holds the node. 3. All corner nodes (gray in Fig. 1) are dropped from consideration. For an elliptic operator of the form (2) with c12 = 0, it turns out that these values do not contribute to any of the spectral derivatives on the interior nodes, which means that the method without corner nodes is mathematically equivalent to the method with corner nodes, see [5, Sec. 2.1] for details. When c12 = 0, one must in order to drop the corner nodes include an extrapolation operator when evaluating the terms involving the spectral representation of the mixed derivative ∂ 2 u/∂x1 ∂x2 . This may lead to a slight drop in the order of convergence, but the difference is hardly noticeable in practice, and the exclusion of corner nodes greatly simplifies the implementation of the method. Since we exclude the corner nodes from consideration, the total number of nodes   in the grid equals N = (p − 2) p n1 n2 + n1 + n2 ≈ p2 n1 n2 . The discretization

134

T. Babb et al.

procedure described then results in an N × N matrix A. For a node i, the value of A(i, :)u depends on what type of node i is: ⎧ ⎪ ⎨ [Au](x i ) for any interior (red) node, A(i, :)u ≈ 0 for any edge node (blue) not on Γ, ⎪ ⎩ ∂u/∂n for any edge node (blue) on Γ. This matrix A can be used to solve BVPs with a variety of different boundary conditions, including Dirichlet, Neumann, Robin, and periodic [12]. In many situations, a simple uniform mesh of the type shown in Fig. 1 is not optimal, since the regularity in the solution may vary greatly, due to corner singularities, localized loads, etc. The HPS method can easily be adapted to handle local refinement. The essential difficulty that arises is that when boxes of different sizes are joined, the collocation nodes along the joint boundary will not align. It is demonstrated in [1, 5] that this difficulty can stably and efficiently be handled by incorporating local interpolation operators.

2.2 A Hierarchical Direct Solver A key observation in previous work on the HPS method is that the sparse linear system that results from the discretization technique described in Sect. 2.1 is particularly well suited for direct solvers, such as the well-known multifrontal solvers that compute an LU-factorization of a sparse matrix. The key is to minimize fill-in by using a so called nested dissection ordering [4, 6]. Such direct solvers are very powerful in a situation where a sequence of linear systems with the same coefficient matrix needs to be solved, since each solve is very fast once the coefficient matrix has been factorized. This is precisely the environment under consideration here. The particular advantage of combining the multidomain spectral collocation discretization described in Sect. 2.1 is that the time required for factorizing the matrix is independent of the local discretization order. As we will see in the numerical experiments, this enables us to attain both very high accuracy, and very high computational efficiency. Remark 1 (General Domains) For simplicity we restrict attention to rectangular domains in this chapter. The extension to domains that can be mapped to a union of rectangles via smooth coordinate maps is relatively straight-forward, since the method can handle variable coefficient operators [12, Sec. 6.4]. Some care must be exercised since singularities may arise at intersections of parameter maps, which may require local refinement to maintain high accuracy. The direct solver described exactly mimics the classical nested dissection method, and has the same asymptotic complexity of O(N 1.5 ) for the “build” (or “factorization”) stage, and then O(N log N) cost for solving a system once the

HPS Accelerated Solvers: Part I, Numerical Experiments

135

coefficient matrix has been factorized. Storage requirements are also O(N log N). A more precise analysis of the complexity that takes into account the dependence on the order p of the local discretization shows [1] that Tbuild ∼ N p4 + N 1.5 , and Tsolve ∼ N p2 + N log N.

3 Time-Stepping Methods For high-order time-stepping of (1), we use the so called Explicit, Singly Diagonally Implicit Runge–Kutta (ESDIRK) methods. These methods have a Butcher diagram with a constant diagonal γ and are of the form 0 2γ c3 .. . cs−1 1

0 γ a3,1 .. . as−1,1 b1 b1

γ a3,2 .. . as−1,2 b2 b2

γ ..

.

..

as−1,3 b3 b3

··· ··· ···

. γ bs−1 bs−1

γ γ

ESDIRK methods offer the advantages of stiff accuracy and L-stability. They are particularly attractive when used in conjunction with direct solvers since the elliptic solve required in each stage involves the same coefficient matrix (I − hγ L), where h is the time-step. In general we split the right hand side of (1) into a stiff part, F [1] , that will be treated implicitly using ESDIRK methods, and a part, F [2] , that will be treated ˆ and b). ˆ Precisely we will use the explicitly (with a Butcher table denoted c, ˆ A, Additive Runge–Kutta (ARK) methods by Carpenter and Kennedy [11], of order 3, 4 and 5. We may choose to formulate the Runge–Kutta method in terms of either solving for slopes or solving for stage solutions. We denote these the ki formulation and the ui formulation, respectively. When solving for slopes the stage computation is kin = F [1] (tn + ci Δt, un + Δt

s 

aij kjn + Δt

j =1

lin = F [2] (tn + ci Δt, un + Δt

s  j =1

s 

aˆ ij ljn ), i = 1, . . . , s,

(3)

aˆ ij ljn ), i = 1, . . . , s.

(4)

j =1

aij kjn + Δt

s  j =1

Note that the explicit nature of (4) is encoded in the fact that the elements on the diagonal and above in Aˆ are zero. Once the slopes have been computed the solution

136

T. Babb et al.

at the next time-step is assembled as un+1 = un + Δt

s 

bj kjn + Δt

j =1

s 

bˆj ljn .

(5)

j =1

If the method is instead formulated in terms of solving for the stage solutions the implicit solves take the form uni

= u + Δt n

s  

 aij F [1] (tn + cj Δt, unj ) + aˆ ij F [2] (tn + cj Δt, unj ) ,

j =1

and the explicit update for un+1 is given by un+1 = un + Δt

s 

bj (F [1] (tn + cj Δt, unj ) + F [2] (tn + cj Δt, unj )).

j =1

The two formulations are algebraically equivalent but offer different advantages. For example, when working with the slopes we do not observe (see experiments presented in the second part of this paper) any order reduction due to time-dependent boundary conditions (see e.g. the analysis by Rosales et al. [14]). On the other hand and as discussed in some detail below, in solving for the slopes the HPS framework requires an additional step to enforce continuity. We note that it is generally preferred to solve for the slopes when implementing implicit Runge–Kutta methods, particularly when solving very stiff problems where the influence of roundoff (or solver tolerance) errors can be magnified by the Lipschitz constant when solving for the stages directly. Remark 2 The HPS method for elliptic solves was previously used in [10], which considered a linear hyperbolic equation ∂u = Lu(x, t), x ∈ Ω, t > 0, ∂t where L is a skew-Hermitian operator. The evolution of the numerical solution can be performed by approximating the propagator exp(τ L) : L2 (Ω) → L2 (Ω) via a rational approximation exp(τ L) ≈

M 

bm (τ L − αm )−1 .

m=−M

If application of (τ L − αm )−1 to the current solution can be reduced to the solution of an elliptic-type PDE it is straightforward to apply the HPS scheme to each term in the approximation. A drawback with this approach is that multiple operators must

HPS Accelerated Solvers: Part I, Numerical Experiments

137

be formed and it is also slightly more convenient to time step non-linear equations using the Runge–Kutta methods we use here. There are two modifications to the HPS algorithm that are necessitated by the use of ARK time integrators, we discuss these in the next two subsections.

3.1 Neumann Data Correction in the Slope Formulation In the HPS algorithm the PDE is enforced on interior nodes and continuity of the normal derivative is enforced on the leaf boundary. Now, due to the structure of the update formula (5), if at some time un has an error component in the null space of the operator that is used to solve for a slope ki , then this will remain throughout the solution process. Although this does not affect the stability of the method it may result in loss of relative accuracy as the solution evolves. As a concrete example consider the heat equation ut = uxx , x ∈ [0, 2], t > 0,

(6)

with the initial data u(x, 0) = 1 − |x − 1|, and with homogenous Dirichlet boundary conditions. We discretize this on two leaves which we denote by α and β. Now in the ki formulation, we solve several PDEs for the ki values and update the solution as un+1 = un + Δt

s 

bj kjn .

j =1

Here, even though the individual slopes have continuous derivatives the kink in un will be propagated to un+1 . In this particular example we would end up with the incorrect steady state solution u(x, t) = 1 − |x − 1|. Fortunately, this can easily be mitigated by adding a consistent penalization of the jump in the derivative of the solution during the merging of two leaves (for details see Section 4 in [1]). That is, if we denote the jump by [[·]] we replace the condition 0 = [[T k + hk ]] where T k is the derivative from the homogenous part and hk is the derivative for the particular solution (of the slope) by the condition [[T k + hk − Δt −1 hu ]] = 0. In comparison to [1] we get the slightly modified merge formula  1 u,α β −1  β k,β u,β  ki,3 = Tα3,3 − T3,3 (h3 − h3 ) , T3,2 ki,2 − Tα3,1 ki,1 + h3 − hk,α 3 − Δt

138

T. Babb et al.

along with the modified equation for the fluxes of the particular solution on the parent box :

9 v1 v2

⎞9 ⎛9 : 9 : :  −1  Tα1,1 0 Tα1,3  α k β β T3,3 − T3,3 − Tα3,1  T3,2 ]⎠ i,1 + = ⎝ + β β ki,2 0 T2,2 T2,3 9

hk,α 1 k,β h2

:

9 +

Tα1,3 β T2,3

:



β

Tα3,3 − T3,3

−1 

β

h3 − hα3 −

1 u,α u,β  (h − h3 ) . Δt 3

Due to space we must refer to [1] for a detailed discussion of these equations. Briefly, hk,α and hk,β above denote the spectral derivative on each child’s boundary for the particular solution to the PDE for ki and are already present in [1]. However, hu,α and hu,β , which denote the spectral derivative of un on the boundary from each child box, are new additions. The above initial data is of course extreme but we note that the problem persists for any non-polynomial initial data with the size of the (stationary) error depending on resolution of the simulation. We further note that the described penalization removes this problem without affecting the accuracy or performance of the overall algorithm. Remark 3 Although for linear constant coefficient PDE it may be possible to project the initial data in a way so that interior affine functions do not cause the difficulty above, for greater generality, we have chosen to enforce the extra penalization throughout the time evolution. Remark 4 When utilizing the ui formulation in a purely implicit problem we do not encounter the difficulty described above. This is because we enforce continuity of the derivative in uns when solving (I − Δtγ L)uns = un + ΔtL

s−1 

s−1   asj unj + Δt asj g(x, tn + cj Δt),

j =1

j =1

followed by the update un+1 = uns .

3.2 Enforcing Continuity in the Explicit Stage The second modification is to the first explicit stage in the ki formulation. Solving a problem with no forcing this stage is simply k1n = L(un ).

HPS Accelerated Solvers: Part I, Numerical Experiments

139

When, for example, L is the Laplacian, we must evaluate it on all nodes on the interior of the physical domain. This includes the nodes on the boundary between two leafs where the spectral approximation to the Laplacian can be different if we use values from different leaves. The seemingly obvious choice, replacing the Laplacian on the leaf boundary by the average, leads to instability. However, stability can be restored if we enforce k1n = L(un ) on the interior of each leaf and continuity of the derivative across each leaf boundary. Algorithmically, this is straightforward as these are the same conditions that are enforced in the regular HPS algorithm, except in this case we simply have an identity equation for k1 on the interior nodes instead of a full PDE. Although it is convenient to enforce continuity of the derivative using the regular HPS algorithm it can be done in a more efficient fashion by forming a separate system of equations involving only data on the leaf boundary nodes. In a single dimension on a discretization with n leafs this reduces the work associated with enforcing continuity of the derivative across leaf boundary nodes from solving n × (p − 1) − 1 equations for n × (p − 1) − 1 unknowns to solving a tridiagonal system of equations n − 1 equations for n − 1 unknowns. In two dimensions the system is slightly different, but if we have n × n leafs with p × p Chebyshev nodes on each leaf then eliminating the explicit equations for the interior nodes reduces the system to (p −2)×2n independent tridiagonal systems of n − 1 equations with n − 1 unknowns for a total of (p − 2) × 2n × (n − 1) equations with (p − 2) × 2n × (n − 1) unknowns. When the ui formulation is used for a fully implicit problem the intermediate stage values still requires us to evaluate Lun , but this quantity only enters through the body load in the intermediate stage PDEs. The explicit first stage in this formulation is simply un1 = un . Furthermore, while we must calculate un+1 = un + ΔtL

s 

 asj unj ,

j =1

this is equivalent to uns since bj = asj and we simply take un+1 = uns . When both explicit and implicit terms are present, we proceed differently. Now, the values of uni look almost identical to the implicit case and we still avoid the problem of an explicit “solve” in un1 , but we also have un+1 = un + Δt

s 

bj (F [1] (tn + cj Δt, unj ) + F [2] (tn + cj Δt, unj ))

j =1

The ESDIRK method has the property that bj = asj , but for the explicit Runge– Kutta method we have bj = aˆ sj . When the explicit operator F [2] does not contain partial derivatives we need not enforce continuity of the derivative and can simply

140

T. Babb et al.

reformulate the method as un+1 = uns + Δt

s  (asj − aˆ sj )F [2] (tn + cj Δt, unj ) j =1

4 Boundary Conditions The above description for Runge–Kutta methods does not address how to impose boundary conditions for a system of ODEs resulting from a discretization of a PDE. In particular, the different formulations incorporate boundary conditions in slightly different ways. In this work we consider Dirichlet, Neumann, and periodic boundary conditions. For periodic boundary conditions the intermediate stage boundary conditions are enforced to be periodic for both formulations. As the ki stage values are approximations to the time derivative of u, the imposed Dirichlet boundary conditions for x ∈ Γ are kin = ut (x, tn + ci Δt). When solving for ui one may attempt to enforce boundary conditions using ui = u(x, t + ci Δt), x ∈ Γ . However, as demonstrated in part two of this series and discussed in detail in [14], this results in order reduction for time dependent boundary conditions. In the HPS algorithm, Neumann or Robin boundary conditions are mapped to Dirichlet boundary conditions using the linear Dirichlet to Neumann operator as discussed for example in [1].

References 1. Babb, T., Gillman, A., Hao, S., Martinsson, P.: An accelerated Poisson solver based on multidomain spectral discretization. BIT Numer. Math. 58, 851–879 (2018) 2. Babb, T., Martinsson, P.-G., Appelö, D.: HPS accelerated spectral solvers for time dependent problems (2018). arXiv:1811.04555 3. Canuto, C., Hussaini, M.Y., Quarteroni, A., Zang, T.A.: Spectral Methods: Evolution to Complex Geometries and Applications to Fluid Dynamics. Springer, Berlin (2007) 4. Duff, I., Erisman, A., Reid, J.: Direct Methods for Sparse Matrices. Oxford University Press, Oxford (1989) 5. Geldermans, P., Gillman, A.: An adaptive high order direct solution technique for elliptic boundary value problems. SIAM J. Sci. Comput. 41(1), A292–A315 (2019). arXiv:1710.08787 6. George, A.: Nested dissection of a regular finite element mesh. SIAM J. Numer. Anal. 10, 345–363 (1973) 7. Gillman, A., Martinsson, P.: A direct solver with O(N) complexity for variable coefficient elliptic PDEs discretized via a high-order composite spectral collocation method. SIAM J. Sci. Comput. 36, A2023–A2046 (2014). arXiv:1307.2665 8. Gillman, A., Barnett, A., Martinsson, P.-G.: A spectrally accurate direct solution technique for frequency-domain scattering problems with variable media. BIT Numer. Math. 55, 141–170 (2015)

HPS Accelerated Solvers: Part I, Numerical Experiments

141

9. Hao, S., Martinsson, P.: A direct solver for elliptic PDEs in three dimensions based on hierarchical merging of Poincaré-Steklov operators. J. Comput. Appl. Math. 308, 419–434 (2016) 10. Haut, T., Babb, T., Martinsson, P., Wingate, B.: A high-order scheme for solving wave propagation problems via the direct construction of an approximate time-evolution operator. IMA J. Numer. Anal. 36, 688–716 (2016) 11. Kennedy, C., Carpenter, M.: Additive Runge–Kutta schemes for convection-diffusion-reaction equations. Appl. Numer. Math. 44, 139–181 (2003) 12. Martinsson, P.: A direct solver for variable coefficient elliptic PDEs discretized via a composite spectral collocation method. J. Comput. Phys. 242, 460–479 (2013) 13. Martinsson, P.: The hierarchical Poincaré-Steklov (HPS) solver for elliptic PDEs: a tutorial (2015). arXiv:1506.01308 14. Rosales, R., Seibold, B., Shirokoff, D., Zhou, D.: Order reduction in high-order Runge–Kutta methods for initial boundary value problems (2017). arXiv:1712.00897 15. Trefethen, L.: Spectral Methods in Matlab. SIAM, Philadelphia (2000)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

On the Use of Hermite Functions for the Vlasov–Poisson System Lorella Fatone, Daniele Funaro, and Gianmarco Manzini

1 Introduction A semi-Lagrangian spectral method has been proposed in [8] for the numerical approximation of the nonrelativistic Vlasov–Poisson equations, which describe the dynamics of a collisionless plasma of charged particles, coupled under the effect of their own electric field. We assume for simplicity that the development of the plasma is only due to electrons. Moreover, we just treat the case of a 1D-1V distribution function, defined in a phase space consisting of the two onedimensional independent variables x (space) and v (velocity). The approximation introduced in [8] has been initially developed and tested on Fourier-Fourier periodic discretizations, for both variables in the phase space. In the successive paper [9], the approximation in the variable v has been approached with the help of Hermite functions, i.e., Hermite polynomials multiplied by the Gaussian weight exp (−v 2 ). Semi-Lagrangian methods for plasma physics calculations were originally proposed in [5, 18] and more recently in [6, 15, 16]. By this approach, at different times, the solution is approximated at the nodes of a Cartesian grid covering the space-

L. Fatone Dipartimento di Matematica, Università degli Studi di Camerino, Camerino, Italy e-mail: [email protected] D. Funaro Dipartimento di Scienze Chimiche e Geologiche, Università degli Studi di Modena e Reggio Emilia, Modena, Italy e-mail: [email protected] G. Manzini () Group T-5, Applied Mathematics and Plasma Physics, Theoretical Division, Los Alamos National Laboratory, Los Alamos, NM, USA e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_10

143

144

L. Fatone et al.

velocity domain. The solution at each space-velocity node is traced back along the characteristic curve originating backward from that node. In [8] a high-order Taylor expansion of the characteristic curves is used to trace back the solution in time, which is then approximated by spectral interpolation. Such a method guarantees the conservation of the main physical quantities (charge, mass, and momentum). The first attempt in using Hermite polynomials to solve the Vlasov equation dates back to the work [10], where the Hermite basis is used in the velocity variable to describe a plasma in a physical state near the thermodynamic equilibrium. Within this approach, exact discrete conservation laws can be constructed [7, 13, 14, 20, 21]. The weight function of the Hermite basis can be generalized by introducing a parameter α in such a way that it becomes exp(−α 2 v 2 ). A proper choice of this parameter can significantly improve the convergence [2, 3, 19]. This fact was also confirmed in earlier works on plasmas physics based on Hermite spectral methods (see [11, 17] and more recently [4]). The paper is organized as follows. In Sect. 2, we present the continuous model, i.e., the 1D-1V Vlasov equation. In Sect. 3, we introduce the spectral approximation in the phase space. In Sect. 4, we present the semi-Lagrangian schemes based on an approximation of the characteristic curves coupled with a second-order backward differentiation formula (BDF). In Sect. 5, we numerically assess the performance of the method for a standard test case, and we show how the solution’s behavior can be affected by the choice of a certain parameter β, acting on the location of Hermite weight function.

2 The Continuous Model We deal with the 1D-1V Vlasov equation defined in the domain  = x × R, with x ⊆ R. The unknown f = f (t, x, v) denotes the probability of finding negative charged particles at the location x with velocity v. This is solution of the problem ∂f ∂f ∂f +v − E(t, x) = 0, ∂t ∂x ∂v

t ∈ (0, T ], x ∈ x , v ∈ R.

(1)

At time t = 0 we have the initial distribution f (0, x, v) = f¯(x, v). The problem is nonlinear, since the electric field E is coupled with f . Indeed, we set ∂E (t, x) = 1 − ρ(t, x) = 1 − ∂x

 R

f (t, x, v)dv,

(2)

where ρ denotes the electron charge density. System (1)–(2) in the unknowns f and E is a simplification of the Vlasov–Poisson equations in two or three dimensional space domains. Uniqueness of the solution is ensured by imposing that 

 E(t, x)dx = 0, x

ρ(t, x)dx = |x |,

which implies that x

(3)

On the Use of Hermite Functions for the Vlasov–Poisson System

145

where |x | is the size of x . We assume periodic boundary conditions in the variable x and a suitable exponential decay at infinity for the variable v. After integration and by using the boundary constraints, we obtain the conservation of mass  d f (t, x, v) dx dv = 0. (4) dt  When f and E are smooth enough, for a sufficiently small δ > 0, the local system of characteristics associated with (1) is given by the curves (X(τ ), V (τ )) solving dX = −V (τ ), dτ

dV = E(τ, X(τ )), dτ

τ ∈]t − δ, t + δ[,

(5)

with the condition that (X(t), V (t)) = (x, v) when τ = t. With this setting we have in mind that for τ > 0 we proceed backward. Under suitable regularity assumptions, there exists a unique solution of the Vlasov–Poisson problem (1)–(2) which is formally obtained by propagating the initial condition along the characteristic curves described by (5), i.e. we have f (t, x, v) = f¯(X(t), V (t)),

(6)

where we recall that f¯ is the initial datum. By using the first-order approximation X(τ ) = x − v(τ − t),

V (τ ) = v + E(t, x)(τ − t),

(7)

the Vlasov equation is satisfied up to an error decaying as |τ − t|, for τ tending to t.

3 Phase-Space Discretization We briefly recall the construction of the approximation method proposed in [8]. At each point of a given grid, the new value of the discrete solution is set up to be equal to the value obtained by going backward, by a suitably small amount, along the local characteristic lines. The algorithm follows from a Taylor expansion of arbitrary order, where the derivatives in the variable x and v are carried out with spectral accuracy. In particular, for the variable x we consider the domain x = [0, 2π[. Given the positive integer N, we have the equispaced nodes xi = 2πi/N , i = 0, 1, . . . , N − 1. Regarding the direction v, when M is a given positive integer, the nodes vj , j = 0, 1, . . . , M − 1, are the zeros of HM , which is the Hermite polynomial of degree M. We introduce the polynomial Lagrangian basis functions for the x and v variables, that are Bi(N) (xn ) = δin and Bj(M) (vm ) = δj m , where δij is the usual Kronecker symbol. We recall that Hermite functions are obtained from Hermite

146

L. Fatone et al.

polynomials after multiplication by the weight ω(v) = e−v . We also define the discrete spaces 2

  (N) XN = span Bi

i=0,1,...,N−1

  (N) (M) YN,M = span Bi Bj ω i=0,1,...,N−1 .

,

(8)

j=0,1,...,M−1

Any function fN,M that belongs to YN,M can be represented as fN,M (x, v) =

N−1  M−1 

cij Bi(N) (x) Bj(M) (v) ω(v),

(9)

i=0 j =0

where the coefficients of such an expansion are given by cij = fN,M (xi , vj ). (N,s) (M,s) (N) In the following, the matrices dni and dmj denote the s-th derivative of Bi (M)

evaluated at point xn and (Bj (N,s) dni

ω) evaluated at point vm

d s Bi(N) = (xn ) dx s

(M,s) dmj

and

=

  d s Bj(M) ω dv s

(10)

(vm ).

(N,0) (M,0) = δni , dmj = δmj . As a special case, we set dni Now, let us assume that the one-dimensional function EN ∈ XN is known. Given

t > 0, by taking τ = t − t in formula (7), we define the new set of points x˜nm = xn − vm t and v˜nm = vm + EN (xn ) t. To evaluate a function fN,M ∈ YN,M at the new points (x˜nm , v˜nm ) through the coefficients cij , we use a Taylor expansion in time. By omitting the terms in t of order higher than one, we get (N)

Bi

  (M) (x˜nm ) Bj ω (v˜nm ) ≈ (N,1) (M,1) δin δj m ω(vm ) − vm t δj m dni ω(vm ) + EN (xn ) t δin dmj .

(11)

By substituting (11) in (9), we obtain the approximation fN,M (x˜nm , v˜nm ) =

N−1  M−1 

cij Bi(N) (x˜nm ) Bj(M) (v˜nm ) ω(v˜nm )

i=0 j =0

≈ cnm ω(vm ) − vm ω(vm ) t

N−1 

(N,1)

dni

cim + EN (xn ) t

i=0

M−1 

(M,1)

dmj

cnj ,

j =0

(12) which is the main building block for more advanced schemes.

On the Use of Hermite Functions for the Vlasov–Poisson System

147

4 Discretization of the Vlasov Equation Given the time instants t k = k t = k T /K for any integer k = 0, 1, . . . , K, we consider the approximation of the unknowns f and E of problem (1)–(2), given by     (k) (k) fN,M (x, v), EN (x)  f (t k , x, v), E(t k , x) ,

x ∈ x , v ∈ R,

(k)

(13)

(k)

where the function fN,M belongs to YN,M and the function EN belongs to XN . Concerning the density function, we define 

(k)

ρN (x) =

(k)

v

fN,M (x, v) dv  ρ(t k , x).

(14)

(k)

Hence, at any time step k, we express fN,M in the following way (k)

fN,M (x, v) =

N−1  M−1 

(k)

(N)

cij Bi

(M)

(x) Bj

(15)

(v)ω(v),

i=0 j =0 (k)

(k)

(0)

where cij = fN,M (xi , vj ). At time t = 0, we use the initial condition cij f (0, xi , vj ) = f¯(xi , vj ). (k) Suppose that EN is given at step k. According to [8], we write (k) (x) = − EN

N/2 "  1 ! (k) aˆ n sin(nx) − bˆn(k) cos(nx) , n

=

(16)

n=1

where the discrete Fourier coefficients aˆ n(k) and bˆn(k), n = 1, 2, . . . , N/2, are suitably (k) related to those of ρN . By taking τ = t − t in (7), we define x˜nm = xn − vm t and v˜nm = (k) vm + EN (xn ) t. The distribution function f is expected to remain constant along the characteristics. The most straightforward discretization method is obtained by advancing the coefficients according to the approximation (k+1) (k) (xn , vm ) ≈ fN,M (x˜nm , v˜nm ). fN,M

(17)

(k+1) This states that the value of fN,M , at the grid points and time step (k + 1) t, is assumed to correspond to the previous value at time k t, recovered by going (k+1) backwards along the characteristics. To compute v˜nm , we should use EN (xn ) (k) instead of EN (xn ). However, the distance between these two quantities is of the order of t, so that the replacement has no practical effects on the accuracy of firstorder methods. Between each step k and the successive one, we need to update the

148

L. Fatone et al.

electric field. This can be done by using the Gaussian quadrature formula in (14), so obtaining (k) (xi ) = ρN

M−1  j =0

M−1  1 1 (k) fN,M c(k) wj , (xi , vj ) wj = ω(vj ) ω(vj ) ij

(18)

j =0

where wj , for j = 1, . . . , M − 1, are the quadrature weights. Afterwards, in order (k+1) to compute the new point-values EN (xn ) of the electric field, it is necessary to (k) integrate ρN . By using approximation (12) in (17), we end up with the first-order explicit scheme of Euler type: (k+1) (k) cnm = cnm + t (k) nm ,

(19)

where (k) nm = −vm

N−1 

(N,1) (k) (k) dni cim + EN (xn )

M−1 

(M,1) (k) dmj cnj

j =0

i=0

1 . ω(vm )

(20)

The parameter t must satisfy a suitable CFL condition, which is obtained by requiring that the point (x˜nm , v˜nm ) falls inside the box ]xn−1 , xn+1 [×]vm−1 , vm+1 [. A straightforward way to increase the time accuracy is to use a multistep discretization scheme as the second-order accurate two-step BDF scheme. We have (k+1) fN,M (xn , vm ) ≈

4 (k) 1 (k−1) ˜ fN,M (x˜nm , v˜nm ) − fN,M (x˜nm , v˜˜nm ), 3 3

(21)

where (x˜nm , v˜nm ) is the point obtained from (xn , vm ) going back of one step t along the characteristic lines. Similarly, the point (x˜˜nm , v˜˜nm ) is obtained by going two steps back along the characteristic lines, i.e., by using 2 t instead of t when computing x˜nm and v˜nm . Despite the fact that a BDF scheme is commonly presented as an implicit technique, in our context (f constant along the characteristics) it assumes the form of an explicit method. In terms of the coefficients, we end up with the scheme  1  4  (k) (k−1) + 2 t (k−1) cnm + t (k) cnm nm − nm 3 3 ⎡ N−1  (N,1) (k) 4 (k) 1 (k−1) 2 ⎣ (k−1) dni (2cim − cim ) = cnm − cnm + t −vm 3 3 3

(k+1) cnm =

i=0

(k) (xn ) +EN

M−1  j =0

⎤ 1 (M,1) (k) (k−1) ⎦. dmj (2cnj − cnj ) ω(vm )

(22)

On the Use of Hermite Functions for the Vlasov–Poisson System

149

From theoretical considerations and the experiments in [8], it turns out that the above method is actually second-order accurate in t. Higher order schemes can be obtained with similar principles. All the above schemes guarantee mass conservation (see (4) for the continuous case), which is a crucial physical property. For practical purposes, it is advisable to make the change of variable f (t, x, v) = p(t, x, v) exp(−v 2 ) in the Vlasov equation, so obtaining  ∂p ∂p ∂p +v − E(t, x) − 2vp = 0, t ∈ (0, T ], x ∈ x , v ∈ R. ∂t ∂x ∂v

(23)

(k)

At time step k, the function p(t k , x, v) is approximated by a function pN,M (x, v) in such a way that pN,M e−v belongs to the finite dimensional space YN,M . A generalization consists in introducing a real parameter α and assuming that the weight function is ω(v) = exp(−α 2 v 2 ). The approximation scheme can be easily adjusted by modifying nodes and weights of the Gaussian formula, through a multiplication by suitable constants. The difficulty in the implementation is practically the same, but, as observed in [9], the results are quite sensitive to the variation of α. 2

(k)

5 Numerical Experiments The numerical scheme here proposed is validated in the standard two-stream instability benchmark test. We consider the Vlasov–Poisson problem (1)–(2) where we set x = [0, 4π[, v = [−5, 5]. The initial solution is given by f¯(x, v) =

" 1 ! √ GR (v) + GL (v) (1 + # cos (κx)), 2a 2π

(24)

where GR (v) = e−α (v−β) and GR (v) = e−α (v+β) are two Gaussians centered √ symmetrically at the points v = ±β. The parameters for (24) are: a = 1/ 8, # = 10−3, κ = 0.5, α = α¯ = 2, β = β¯ = 1. In all the experiments that follow, we integrate up to time T = 30 using the second-order BDF scheme with a suitably small time step, in order to guarantee stability and a good accuracy. In this way we can concentrate our attention to the spectral approximation in the variable x and v. A study of the convergence rate in time of the proposed numerical scheme can be found in [8]. First of all, in Fig. 1 we show the results at time T = 30 of the solution recovered by the Fourier-Fourier method, by choosing N = 25 , M = 26 and time step equal to t = 0.00125. This will be the referring figure for the successive comparisons. Besides we show (k) the corresponding time evolution of |aˆ 1 |, the first Fourier mode of the electric (k) field EN in (16). The behavior of this last quantity is predicted by theoretical 2

2

2

2

150

L. Fatone et al.

Fig. 1 Two-stream instability test: approximated distribution function at time T = 30 obtained by using the Fourier-Fourier method with N = 25 , M = 26 , t = 0.00125, and the corresponding (k) , i.e. |aˆ 1(k) | in (16) time evolution of the first Fourier mode of the electric field EN

considerations, and the slope of the “segment” starting at T = 15 agrees with the expectancy [1, Chapter 5]. As done in [9], we perform a series of experiments using less degrees of freedom than those actually necessary to resolve accurately the equation. In practice, we set N = M = 24 . In this way, we could for instance detect what happens by varying the parameters α and β. Of course, if we increase the number of degrees of freedom, the numerical solution improves and cannot be distinguished from the referring one shown in Fig. 1. The purpose in [9] was to check what happens by varying the parameter α in the Hermite weight exp (−α 2 v 2 ). The conclusions are that the approximate solution is very sensitive to the choice of α and that there are values of α that perform better than others. In general these values are those belonging to a neighbourhood of α = 1. Moreover, in [9], we note that keeping α constantly equal to the value that better fits the initial datum (i.e. α = α¯ = 2 for (24)) may create instability as time increases. For such motivations, since at the moment a practical algorithm able to vary α in a dynamical way during the computations is not available, in the numerical experiments that follow we fix α = 1, while play with β. Due to the particular initial condition, we adopt a two-species decomposition of the Vlasov equation, where the distribution function is given by the sum of two electron distribution functions, i.e., f = fR + fL . These distribution functions refer to the two initial electron distributions, so that fR = pR GR and fL = pL GL , where pL and pR are given polynomials. We consider the two systems of electrons described by the distribution functions fL and fR at the initial time as distinct plasma species that maintain their diversity throughout the whole numerical simulation. Therefore, we can split the Vlasov equation into two equations that are still of Vlasov type and are solvable independently, although they are coupled through the same electric field, which depends on the total charge density. This amounts to approximate two independent equations of the same type of that given

On the Use of Hermite Functions for the Vlasov–Poisson System

151

Fig. 2 Two-stream instability test: approximated distribution function at time T = 30 obtained by using the Fourier–Hermite method with N = M = 24 , t = 0.01, α = 1 (left panel) and the (k) , i.e. |aˆ 1(k) | in (16) corresponding time evolution of the first Fourier mode of the electric field EN (right panel) when β = 0.5 (top), β = 1 (center) and β = 1.5 (bottom)

152

L. Fatone et al.

in (23), respectively shifted by ±β, i.e.  ∂pR ∂pR ∂pR + (v − β) − E(t, x) − 2α 2 (v − β)pR = 0, ∂t ∂x ∂v  ∂pL ∂pL ∂pL + (v + β) − E(t, x) − 2α 2 (v + β)pL = 0. ∂t ∂x ∂v

(25) (26)

The two unknowns are then coupled through the density function as in (2). The plots of Fig. 2 show the numerical distribution function at time T = 30 obtained by using the Fourier–Hermite method with N = M = 24 , t = 0.01, α = 1 and different values of the parameter β (i.e. β = 0.5, β = 1 and β = 1.5), together with the corresponding time evolution of the (log of the) first Fourier mode (k) (k) of the electric field EN , i.e. |aˆ 1 | in (16). The distribution functions presented in the left column of Fig. 2 are visibly and significantly different depending on β, while the first Fourier mode of the electric field shown in the right column seems to be less affected. These differences practically confirm that the choice of the Hermite weight functions ω(v) = exp(−α 2 (v ± β)2 ) is a crucial aspect of the method (see also [11, 12, 17, 22]). This conclusion is heuristic. Unfortunately, there is no space enough for a deeper quantitative analysis in these pages. The question deserves however further investigation. Moreover, it would be advisable to develop appropriate algorithms allowing for the automatic adjustment of both parameters α and β during the time advancing procedure, in order to optimize the performance.

References 1. Bittencourt, J.A.: Fundamentals of Plasma Physics. Springer, New York (2004) 2. Boyd, J.P.: The rate of convergence of Hermite function series. Math. Comput. 35, 1309–1316 (1980) 3. Boyd, J.P.: Asymptotic coefficients of Hermite function series. J. Comput. Phys. 54, 382–410 (1984) 4. Camporeale, E., et al. On the velocity space discretization for the Vlasov–Poisson system: comparison between implicit Hermite spectral and Particle-in-Cell methods. Comput. Phys. Commun. 198, 47–58 (2016) 5. Cheng, C.Z., Knorr, G.: The integration of the Vlasov equation in configuration space. J. Comput. Phys. 22(3), 330–351 (1976) 6. Crouseilles, N., Respaud, T., Sonnendrücker, E.: A forward semi-Lagrangian method for the numerical solution of the Vlasov equation. Comput. Phys. Commun. 180(10), 1730–1745 (2009) 7. Delzanno, G.L.: Multi-dimensional, fully-implicit, spectral method for the Vlasov–Maxwell equations with exact conservation laws in discrete form. J. Comput. Phys. 301, 338–356 (2015) 8. Fatone, D., Funaro, L., Manzini, G.: Arbitrary-order time-accurate semi-Lagrangian spectral approximations of the Vlasov–Poisson system. J. Comput. Phys. 384, 349–375 (2019)

On the Use of Hermite Functions for the Vlasov–Poisson System

153

9. Fatone, D., Funaro, L., Manzini, G.: A semi-Lagrangian spectral method for the Vlasov– Poisson system based on Fourier, Legendre and Hermite polynomials. Commun. Appl. Math. Comput. 1, 333–360 (2019) 10. Grad, H.: On the kinetic theory of rarefied gases. Commun. Pure Appl. Math. 2(4), 331–407 (1949) 11. Holloway, J.P.: Spectral velocity discretizations for the Vlasov-Maxwell equations. Transp. Theory Stat. Phys. 25(1), 1–32 (1996) 12. Ma, H., Sun, W., Tang, T.: Hermite spectral methods with a time-dependent scaling for parabolic equations in unbounded domains. SIAM J. Numer. Anal. 43, 58–75 (2005) 13. Manzini, G., et al. A Legendre-Fourier spectral method with exact conservation laws for the Vlasov–Poisson system. J. Comput. Phys. 317, 82–107 (2016) 14. Manzini, G., Funaro, D., Delzanno, G.L.: Convergence of spectral discretizations of the Vlasov–Poisson system. SIAM J. Numer. Anal. 55(5), 2312–2335 (2017) 15. Qiu, J.-M., Christlieb, A.: A conservative high order semi-Lagrangian WENO method for the Vlasov equation. J. Comput. Phys. 229(4), 1130–1149 (2010) 16. Qiu, J.-M., Russo, G.: A high order multidimensional characteristic tracing strategy for the Vlasov–Poisson system. J. Sci. Comput. 71, 414–434 (2017) 17. Schumer, J.W., Holloway, J.P.: Vlasov simulations using velocity-scaled Hermite representations. J. Comput. Phys. 144(2), 626–661 (1998) 18. Sonnendrücker, E., et al.: The semi-Lagrangian method for the numerical resolution of the Vlasov equation. J. Comput. Phys. 149(2), 201–220 (1999) 19. Tang, T.: The Hermite spectral method for Gaussian-type functions. SIAM J. Sci. Comput. 14(3), 594–606 (1993) 20. Vencels, J., et al.: Spectral solver for multi-scale plasma physics simulations with dynamically adaptive number of moments. Proc. Comput. Sci. 51, 1148–1157 (2015) 21. Vencels, J., et al.: SpectralPlasmaSolver: a spectral code for multiscale simulations of collisionless, magnetized plasmas. J. Phys. Conf. Series 719(1), 012022 (2016) 22. Xiang, X.-M., Wang, Z.-Q.: Generalized Hermite approximations and spectral method for partial differential equations in multiple dimensions. J. Sci. Comput. 57, 229–253 (2013)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

HPS Accelerated Spectral Solvers for Time Dependent Problems: Part I, Algorithms Tracy Babb, Per-Gunnar Martinsson, and Daniel Appelö

1 Introduction In this chapter, part two in a two part series, describes a sequence of numerical experiments demonstrating the performance of a highly computationally efficient solver for equations of the form κ

∂u = Lu(x, t) + g(u, x, t), x ∈ Ω, t > 0, ∂t

(1)

with initial data u(x, 0) = u0 (x). Here L is an elliptic operator acting on a fixed domain Ω and f is lower order, possibly nonlinear terms. We take κ to be real or imaginary, allowing for parabolic and Schrödinger type equations. The “Hierarchial Poincaré–Steklov (HPS)” solver has already been demonstrated to be a highly competitive spectrally accurate solver for elliptic problems [1, 4, 7] and has also been used together with a class of exponential integrators [5], to evolve solutions to hyperbolic differential equations. As just mentioned, the focus here is on differential equations in the form (1) whose discretization leads to stiff system of ODE that can beneficially be advanced in time using Explicit, Singly Diagonally Implicit Runge–Kutta (ESDIRK) methods. ESDIRK methods offer the advantages of stiff accuracy and L-stability and are well suited for the HPS algorithm as they only require a single matrix factorization. They are also easily combined with

T. Babb · D. Appelö () University of Colorado, Boulder, CO, USA e-mail: [email protected]; [email protected] P.-G. Martinsson University of Texas, Austin, TX, USA e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_11

155

156

T. Babb et al.

explicit Runge–Kutta method leading to so called Additive Runge–Kutta (ARK) methods [6]. To this end we investigate the stability and accuracy that is obtained when combining high-order time-stepping schemes with the HPS method for solving elliptic equations. We restrict attention to relatively simple geometries (rectangles) but note that the method can without difficulty be generalized to domains that can be expressed as a union of rectangles, possibly mapped via curvilinear smooth parameter maps. The rest of this chapter is organized as follows. In Sect. 2 we present results illustrating that the order reduction phenomena for DIRK methods observed in [8] can be circumvented when formulating the time stepping in terms of slopes (with boundary conditions differentiated in time) rather than formulating it in terms of stage solutions. In Sect. 3 we present numerical results for Schrödingers equation in two dimensions and in Sect. 4 we present numerical results for a nonlinear problem, viscous Burgers’ equation in two dimensions. Finally, in Sect. 5 we summarize and conclude. For a longer description of the method we refer to thee first part of this paper and to [2].

2 Time Dependent Boundary Conditions This section discusses time-dependent boundary conditions within the two different Runge–Kutta formulations. In particular, we investigate the order reduction that has been documented in [8] for implicit Runge–Kutta methods and earlier in [3] for explicit Runge–Kutta methods. In this first experiment, introduced in [8], we solve the heat equation in one dimension ut = uxx + f (t),

x ∈ [0, 2], t > 0.

(2)

We set the initial data, Dirichlet boundary conditions and the forcing f (t) so that exact solution is u(x, t) = cos(t). This example is designed to eliminate the effect of the spatial discretization, with the solution being constant in space and allows for the study of possible order reduction near the boundaries. We use the HPS scheme in space and use 32 leafs with p = 32 Chebyshev nodes per leaf. We apply the third, fourth, and fifth order ESDIRK methods from [6]. We consider solving for the intermediate solutions, or as we refer to it below “the ui formulation” with the boundary condition enforced as uni = cos(tn + ci Δt). We also consider solving for the stages, which we refer to as “the ki formulation” with boundary conditions imposed as kin = − sin(x, tn + ci Δt). Error reduction for time dependent boundary conditions has been studied both in the context of explicit Runge–Kutta methods in e.g. [3] and more recently for implicit Runge–Kutta methods in [8]. In [8] the authors report observed orders of accuracy equal to two (for the solution u) for DIRK methods of order 2, 3, and 4 for

HPS Accelerated Solvers: Part I, Algorithms

157

−8

−6

x 10

x 10

−4

−6

Δt=2 4

2

0

0

0.5

1 x

Δ t = 2−4

3

−5

Δt=2

6

u(x,1) − cos(1)

u(x,Δ t) − cos(Δ t)

Δt=2

1.5

Δ t = 2−5

1

0

2

Δ t = 2−6

2

0

0.5

(a)

1 x

1.5

2

(b)

Fig. 1 The error in solving (2). Results are for a third order ESDIRK. (a) Displays the single step error which converges with fourth order of accuracy. (b) Displays the global error at t = 1 converging at third order. Both errors converge at one order higher than what is expected from the analysis in [8] −9

−8

x 10

x 10 −4

Δt=2

−4

Δt=2

15

−5

−5

Δt=2

6

u(x,1) − cos(1)

u(x,Δ t) − cos(Δ t)

8

−6

Δt=2

4 2

Δt=2

Δ t = 2−6

10 5 0

0 0

0.5

1 x

(a)

1.5

2

0

0.5

1 x

1.5

2

(b)

Fig. 2 The error in solving (2). Results are for a fifth order ESDIRK. (a) Displays the single step error which converges with fourth order of accuracy. (b) Displays the global error at t = 1 converging at third order. Both errors converge at one order higher than what is expected from the analysis in [8] but still lower than expected

the problem (2) discretized with a finite difference method on a fine grid (the spatial errors are zero) using the ui formulation. Figures 1 and 2 show the error for the third and fifth order ESDIRK methods, respectively, as a function of x for a single step and at the final time t = 1. Figure 3 shows the maximum error for the third, fourth, and fifth order methods as a function of time step Δt after a single step and at the final time t = 1. In general, for a method of order p we expect that the single step error decreases as Δt p+1 while the global error decreases as Δt p . However, with time dependent

158

T. Babb et al.

3rd order 4th order 5th order 4th order slope

∞

l∞ Error

10

3rd order 4th order 5th order 3rd order slope

−4

10

l Error

−5

−6

10

−8

−10

10

10

10

−2

10

−1

10

Δt

−2

Δt

(a) 3rd order 4th order 5th order 4th order slope

−6

10

l Error

−8

10

∞

l∞ Error

−1

(b)

3rd order 4th order 5th order 6th order slope

−6

10

10

−8

10

−10

10

−10

10

−12

10

−2

10

−1

10

−2

10

−1

10

Δt

Δt

(c)

(d)

Fig. 3 The maximum error (here denoted l ∞ ) in solving (2) for the third, fourth, and fifth order ESDIRK methods for a sequence of decreasing time steps. (a, c) are errors after one time step and (b, d) are the errors at time t = 1. The top row are for the ui formulation and the bottom row is for the ki formulation. Note that the ki formulation is free of order reduction

boundary conditions implemented as uni = cos(tn + ci Δt) the results in [8] indicate that the rate of convergence will not exceed two for the single step or global error. The results for the third order method (p = 3) displayed in Fig. 1 show that the single step error decreases as Δt p+1 while the global error decreases as Δt p , which is better than the results documented in [8]. However, we still see that a boundary layer appears to be forming, but it is of the same order as the error away from the boundary. The results for the fifth order method (p = 5) displayed in Fig. 2 show that the single step error decreases as Δt 4 while the global error decreases as Δt 3 , which is still better than the results documented in [8]. However, the boundary layer is giving order reduction from Δt p+1 for the single step error and Δt p for the global error. We note that our observations differ from those in [8] but that this possibly can be attributed to the use of a ESDIRK method rather than a DIRK method. We repeat the experiment but now we use the ki formulation for Runge–Kutta methods and for the boundary condition we enforce kin = − sin(tn + ci Δt). The intuition here is that kin is an approximation to ut at time tn + ci Δt and we use the value of ut for the boundary condition of kin . Intuitively we expect that the fact

HPS Accelerated Solvers: Part I, Algorithms

159

that we reduce the index of the system of differential algebraic equation in the ui formulation by differentiating the boundary conditions can restore the design order of accuracy. In the previous examples the Runge–Kutta method introduced an error on the interior while the solution on the boundary was exact. If the error on the boundary is on the same order of magnitude as the error on the interior then the error in uxx is of the correct order, but when the value of u is exact on the boundary it introduces a larger error in uxx . In the ki formulation, for each intermediate stage we find uxx = 0 and then kin = − sin(tn + ci Δt) on the interior and on the boundary. So at a fixed time the solution is constant in x and a boundary layer does not form. Additionally, the error is constant in x at any fixed time and for a method of order p we obtain the expected behavior where the single step error decreases as Δt p+1 and the global error decreases as Δt p . Figure 3 shows the maximum error for the third, fourth, and fifth order methods as a function of time step Δt after a single step and at the final time t = 1. The results show that the methods behave exactly as we expect. The single step error behaves as Δt p+1 for the third and fifth order methods and Δt p+2 for the fourth order method. The fourth order method gives sixth order error in a single step because the exact solution is u(x, t) = cos(t), which has every other derivative equal to zero at t = 0 and for a single step we start at t = 0. The global error behaves as Δt p for each method.

3 Schrödinger Equation Next we consider the Schrödinger equation for u = u(x, y, t) i h¯ ut = −

h¯ 2 Δu + V (x, y)u, t > 0, (x, y) ∈ [xl , xr ] × [yb , yt ], 2M

(3)

u(x, y, 0) = u0 (x, y). Here we nondimensionalize in a way equivalent to setting M = 1, h¯ = 1 in the above equation. We choose the potential to be the harmonic potential V (x, y) =

 1 2 x + y2 . 2

This leads to an exact solution u(x, y, t) = Ae−it e−

(x 2 +y 2 ) 2

,

(4)

,√ π and solve until t = 2π on the domain (x, y) ∈ [−8, 8]2. where we set A = 1/

T. Babb et al.

10-2

10-2

10-4

10-4

10-6 p=4 p=6 p=8 p = 10 p = 12

10-8

10-10

0.3

0.4

0.5

0.6

0.7 0.8 0.9 1

Maximum error

Maximum error

160

10-6 p=4 p=6 p=8 p = 10 p = 12

10-8

10-10

0.3

0.4

Leaf width

0.5

0.6

0.7 0.8 0.9 1

Leaf width

Fig. 4 Error in the Schrödinger equation as a function of leaf size. The exact solution is given in Eq. (4) Table 1 Estimated rates of convergence for different Runge–Kutta methods and different orders of approximation

p ESDIRK3 ESDIRK4 ESDIRK5

4 2.59 1.89 1.84

6 5.73 6.47 4.42

8 7.72 7.82 7.69

10 9.69 9.76 9.71

12 11.47 11.69 11.48

The computational domain is subdivided into nx × ny panels with p × p points on each panel. To begin, we study the order of accuracy with respect to leaf size. To eliminate the effect of time-stepping errors we scale Δt = hp/qRK , where qRK is the order of the Runge–Kutta method. In Fig. 4 we display the errors as a function of the leaf size for p = 4, 6, 8, 10, 12, 16 and for the third and fifth order Runge–Kutta methods (qRK = 3, 5). The rates of convergence are found for all three Runge– Kutta methods and summarized in Table 1. As can be seen from the table, p = 4 appears to converge at second order, while for higher p we generally observe a rate of convergence approaching to p. In this problem the efficiency of the method is limited by the order of the Runge–Kutta methods. However, as our methods are unconditionally stable we may enhance the efficiency by using Richardson extrapolation to achieve a highly accurate solution in time. We solve the same problem, but now we fix p = 12 and take 5 · 2n time steps, with n = 0, 1, . . . , 5. For the third order ESDIRK method we use 60 × 60 leaf boxes. For the fourth order ESDIRK method we use 90×90 leaf boxes. For the fifth order ESDIRK method we use 120×120 leaf boxes. Table 2 shows that we can easily achieve much higher accuracy by using Richardson extrapolation. Finally, we solve a problem without an analytic solution. In this problem the initial data u(x, y, t) = 3 sin(x) sin(y)e−(x

2 +y 2 )

,

HPS Accelerated Solvers: Part I, Algorithms

161

Table 2 Estimated errors at the final time after Richardson extrapolation qRK / extrapolations 0

1

2

3

4

5

6

3 4

1.32 (−1) 1.01 (−2) 1.27 (−4) 1.17 (−5) 6.98 (−8) 8.62 (−10) 7.40 (−6) 2.70 (−4) 6.46 (−6) 1.23 (−7) 2.95 (−10) 1.59 (−11) 3.70 (−14) 1.20 (−11)

5

1.28 (−3) 9.67 (−6) 6.30 (−8) 1.86 (−10) 4.11 (−13) 9.27 (−14) 5.08 (−11)

The notation d(−p) means d · 10−p Table 3 Errors computed against a p and h refined solution p/panels 8 Rate 10 Rate

2 1.11 (0) ∗ 5.87 (−1) ∗

4 1.39 (−1) 3.00 3.16 (−2) 4.21

8 8.74 (−3) 3.99 4.62 (−4) 6.10

16 1.50 (−4) 5.87 6.17 (−6) 6.22

32 2.45 (−6) 5.92 5.21 (−8) 6.89

The errors are maximum errors at the final time t = 4. The notation d(−p) means d · 10−p

interacts with the weak and slightly non-symmetric potential V (x, y) = 1 − e−(x+0.9y) , 4

allowing the solution to reach the boundary where we impose homogenous Dirichlet conditions. We evolve the solution until time t = 4 using p = 8 and 10 and 2, 4, 8, 16 and 32 leaf boxes in each direction of a domain of size 12 × 12. The errors computed against a reference solution with p = 12 and with 32 leaf boxes can be found in Table 3. In Fig. 5 we display snapshots of the magnitude of the solution at the initial time t = 0, the intermediate times t ≈ 1.07, t ≈ 1.68 and at the final time t = 4.0.

4 Burgers’ Equation in Two Dimensions As a first step towards a full blown flow solver we solve Burgers’ equation in two dimensions using the additive Runge–Kutta methods described in the first part of this paper. Precisely, we solve the system ut + u · ∇u = εΔu,

x ∈ [−π, π]2 , t > 0,

(5)

where u = [u(x, y, t), v(x, y, t)]T is the vector containing the velocities in the x and y directions. The first problem we solve uses the initial condition u = 5[−y, x]T exp(−3r 2 ) and the boundary conditions are taken to be no-slip boundary conditions on all sides.

162

T. Babb et al.

Fig. 5 Snapshots of the magnitude of the solution at the initial time (a) t = 0, the intermediate times (b) t ≈ 1.07, (c) t ≈ 1.68 and at the final time (d) t = 4.0

We solve the problem using 24 × 24 leafs, p = 24, ε = 0.005, and the fifth order ARK method found in [6]. We use a time step of k = 1/80 and solve until time tmax = 5. The low viscosity combined with the initial condition produces a rotating flow resembling a vortex that steepens up over time. In Fig. 6 we can see the velocities at times t = 0.5 and t = 1. The fluid rotates and expands out and eventually forms a shock like transition. This creates a sharp flow region with large gradients resulting in a flow that may be difficult to resolve with a low order accurate method. These sharp gradients can be seen in the two vorticity plots in Fig. 6 along with the speed and vorticity plots in Fig. 7. In our second experiment we consider a cross stream of orthogonal flows. We use an initial condition of u = [8y e

−36

 8 y 2

, −8xe

−36

 8 x 2

]T ,

(6)

and time independent boundary conditions that are compatible with the initial data.

-3

-3

-2

-2

-1

-1

0

u at time t= 1

0

1

1

2

2

3

3

-3

-2

-1

0

1

2

3

-3 -3

-2

-1

0

1

2

3

-0.8

-0.8

-0.4

-0.2

0

0.2

0.4

0.6

0.8

-0.6

-1

0

1

2

3

-0.6 -2

-0.4

-0.2

0

0.2

0.4

0.6

0.8

v at time t= 1

-1

-3

-1

-0.6

-0.4

-0.2

0

0.2

-0.8

-2

-1

0

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.8

1

0.4

1

v at time t= 0.5

0.6

2

3

0.6

0.8

1

2

-0.5

0

0.5

1

1.5

2

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

-3

-3

-2

-2

-1

-1

x

0

x

0

1

1

2

2

3

3

Fig. 6 The four plots on the left show the velocities in the x and y directions at times t = 0.5 and t = 1. We can see the fluid rotating and expanding. The two plots in the third column show the vorticity at these times. We see sharp gradients near the edge of the rotating fluid

-3

-2

-1

0

1

2

3

-3

-2

-1

0

1

2

u at time t= 0.5

(x,0,0.5) Vorticity (x,0,1) Vorticity

3

HPS Accelerated Solvers: Part I, Algorithms 163

-0.5

-0.5

0

u at time t= 0.15

0

u at time t= 0.06

0.5

0.5

1

1

-1.5

-1

-0.5

0

0.5

1

1.5

-1.5

-1

-0.5

0

0.5

1

1.5

Fig. 7 The plots in the first column correspond to the rotating shock problem. The second and third columns show the velocity in the x direction and the dilatation at times t = 0.06 and t = 0.15 for the cross flow problem

-1 -1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

-1 -1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

164 T. Babb et al.

HPS Accelerated Solvers: Part I, Algorithms

165

This initial horizontal velocity drops to zero quickly as we approach |y| = 0.5. For |y| < 0.5 the exponential term approaches exp(0) and the velocity behaves like u = 8y. The flow has changed slightly by t = 0.06, but we can see in Fig. 7 the flow is moving to the right for y > 0 and the flow is moving the left for y < 0 and all significant behavior is in |y| < 0.5. A plot of the velocity v would show similar behavior. We also use 24 × 24 leafs, p = 24, # = 0.025, k = 1/200, and tmax = 0.75. We show plots of the horizontal velocity u and the dilatation at time t = 0.06 and t = 0.15. We only show plots before time t = 0.15 when the fluid is hardest to resolve and we observe that after t = 0.15 the cross streams begin to dissipate. This problem contains sharp interfaces inside x ∈ [−0.5, 0.5]2.

5 Conclusion In this two part series we have demonstrated that the spectrally accurate Hierarchial Poincaré–Steklov solver can be easily extended to handle time dependent PDE problems with a parabolic principal part by using ESDIRK methods. We have outlined the advantages of the two possible ways to formulate implicit Runge–Kutta methods within the HPS scheme and demonstrated the capabilities on both linear and non-linear examples. There are many avenues for future work, for example: • Extension of the solvers to compressible and incompressible flows. • Application of the current solvers to inverse and optimal design problems, in particular for problems where changes in parameters do not require new factorizations.

References 1. Babb, T., Gillman, A., Hao, S., Martinsson, P.: An accelerated Poisson solver based on multidomain spectral discretization. BIT Numer. Math. 58, 851–879 (2018) 2. Babb, T., Martinsson, P.-G., Appelö, D.: HPS accelerated spectral solvers for time dependent problems (2018). arXiv:1811.04555 3. Carpenter, M., Gottlieb, D., Abarbanel, S., Don, W.-S.: The theoretical accuracy of Runge– Kutta time discretizations for the initial boundary value problem: a study of the boundary error. SIAM J. Sci. Comput. 16, 1241–1252 (1995) 4. Gillman, A., Martinsson, P.: A direct solver with O(N) complexity for variable coefficient elliptic PDEs discretized via a high-order composite spectral collocation method. SIAM J. Sci. Comput. 36, A2023–A2046 (2014). arXiv:1307.2665 5. Haut, T., Babb, T., Martinsson, P., Wingate, B.: A high-order scheme for solving wave propagation problems via the direct construction of an approximate time-evolution operator. IMA J. Numer. Anal. 36, 688–716 (2016) 6. Kennedy, C., Carpenter, M.: Additive Runge–Kutta schemes for convection-diffusion-reaction equations. Appl. Numer. Math. 44, 139–181 (2003)

166

T. Babb et al.

7. Martinsson, P.: A direct solver for variable coefficient elliptic PDEs discretized via a composite spectral collocation method. J. Comp. Phys. 242, 460–479 (2013) 8. Rosales, R., Seibold, B., Shirokoff, D., Zhou, D.: Order reduction in high-order Runge–Kutta methods for initial boundary value problems (2017). arXiv:1712.00897

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

High-Order Finite Element Methods for Interface Problems: Theory and Implementations Yuanming Xiao, Fangman Zhai, Linbo Zhang, and Weiying Zheng

1 Introduction The interface problems which involve partial differential equations having discontinuous coefficients across certain interfaces are often encountered in fluid dynamics, electromagnetics and materials science. Because of the low global regularity and the irregular geometry of the interface, the standard numerical methods which are efficient for smooth solutions usually lead to loss in accuracy across the interface. For arbitrarily shaped interface , it is known that optimal or nearly optimal convergence rate can be recovered if body-fitted finite element meshes are used, see e.g. [6, 8, 20, 29]. Here, by “body-fitted meshes” we mean an element of the underlying mesh is required to intersect with the interface only through its boundaries (Fig. 1). Unfortunately, when the geometry is complex, this usually leads to a nontrivial interface meshing problem. Therefore, numerous modified finite difference methods based only on simple Cartesian grids have been proposed in the literature. We refer to the immersed boundary method [24], the immersed interface method [17, 18], the ghost fluid method [21], and the references therein. In the Y. Xiao () Department of Mathematics, Nanjing University, Nanjing, China e-mail: [email protected] F. Zhai Department of Applied Mathematics, Nanjing Forestry University, Nanjing, China L. Zhang · W. Zheng State Key Laboratory of Scientific and Engineering Computing, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing, China School of Mathematical Sciences, University of Chinese Academy of Sciences, Beijing, China e-mail: [email protected]; [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_12

167

168

Y. Xiao et al.

Fig. 1 A body-fitted, shape regular mesh

finite element setting, we refer to the work of the immersed finite element method [7, 11, 19], the multiscale finite element method [9], the penalty finite element method [1]. In the past decade, a combination of the extended finite element method (XFEM) with the Nitsche scheme has become a popular discretization method. As the first attempt, an unfitted finite element method was proposed in [13] which can be viewed as a linear and consistent modification of [1]. This approach has motivated a number of works, e.g., the unfitted finite element method [4, 5, 12], the Ghost penalty method [2, 3], the unfitted discontinuous Galerkin methods [22]. Although significant progresses in the error analyses of some methods have been made, the development of high-order accurate unfitted FEMs with rigorous error analysis is still challenging. We refer to the work of [14–16, 22, 27, 28] which claim high order approximations. In [22], an hp-unfitted discontinuous Galerkin method for Problem (1) was considered, and optimal h-convergence for arbitrary p was shown for the two-dimensional case in the energy norm and in the L2 -norm. With an extra flux penalty term applied on the interface, [27] gave better hp a priori error estimates in both two and three dimensions. In [15, 16], an isoparametric finite element method with a high order geometrical approximation of level set domains was presented. The analysis reveals optimal order error bounds with respect to h for the geometry approximation and for the finite element approximation. In [14, 28], various issues related to unfitted methods was addressed, including the dependence of error estimates on the diffusion coefficients, the condition number of the discrete system, and the choice of stabilization parameters. The Nitsche-XFEM can be interpreted as applying interior penalty (IP) methods on the interface, and our method falls into this category. The major step in our variant is an appropriate choice of the mesh and geometry dependent weights in the average (see (6)), which lead to trace and inverse inequalities for possibly degenerated subelements (see (9)). We note that in our approach, the penalization is applied only to the jump of the solution values across the interface (compared with the bilinear form in [27]). The optimal h-convergence rate for arbitrary high-order discretization in the energy and L2 -norm are proved regardless of the dimension. We refer to [14– 16] for the similar estimates with respect to h and [27] for a refined version with respect to both h and p.

High-Order FEMs for Interface Problems

169

Efficient implementations of this method are then discussed in two aspects. We first consider an optimal multigrid solver for the generated linear system. We use the continuous FE space as a “background” subspace, with some smoothing operations added near the interface, to formulate a nested geometrical multigrid method. We prove the optimality of this special multiplicative multigrid method, which means the method converges uniformly with respect to the mesh size, and is independent of the location of the interface relative to the meshes. Since the assembling of the stiffness matrix will require integration over curved surfaces and volumes, we then implement a robust and arbitrarily high order numerical quadrature algorithm by transforming surface and volume integrals into multiple 1-D integrals. The code for the algorithm is freely available in the open source finite element toolbox Parallel Hierarchical Grid (PHG) [26]. We also refer to [23, 25] for different approaches to compute integrals on curved sub-elements and their curved boundaries. The layout of this paper is as follows. In Sect. 2 we introduce the XFE spaces and reformulate the interface problem (1) in DG schemes. The H 1 - and L2 - error estimates of both schemes—which attain the optimal order of the convergence rate in respect to mesh size h—are given. In Sect. 3, we give an optimal multigrid method for the aforementioned DG-XFE schemes. Numerical examples for both two and three dimensions are reported in Sect. 4, to illustrate the high accuracy of the algorithm.

2 XFE and DG Schemes for Interface Problems We consider the following elliptic interface problem for u: Let  = 1 ∪  ∪ 2 be a bounded and convex polygonal or polyhedral domain in Rd , d = 2 or 3, where 1 and 2 are two subdomains of  and are separated by a C 2 -smooth interface  (see Fig. 2 for an illustration of a unit square that contains a circle as an interface), ⎧ ⎪ ⎪ −∇ · (α(x)∇u) = ⎪ ⎨ α(x)∇u = ⎪ [u] = ⎪ ⎪ ⎩ u=

f, gN , gD , 0,

in 1 ∪ 2 , on , on , on ∂.

(1)

Here α(x) = αi , i = 1, 2, is a piecewise constant function on the partition 1 ∪ 2 . Denote by {Th }, a family of conforming, quasi-uniform, and regular partitions of  into triangles and parallelograms/tetrahedrons and parallelepipeds. As K is of regular shape, there is a constant γ0 such that hdK ≤ γ0 |K|,

∀K ∈ Th .

(2)

We define the set of all elements intersected by  as Th = {K ∈ Th : |K ∩ | = 0}. Each Th induces a partition of interface , which we denote by Eh = {eK : eK =

170

Y. Xiao et al.

Fig. 2 Domain  = 1 ∪  ∪ 2 with an unfitted mesh

Ω2

Ω1

K ∩ , K ∈ Th }. For any K ∈ Th , let Ki = K ∩ i denote the part of K in i and ni be the unit outward normal vector on ∂Ki with i = 1, 2. As  is of class C 2 , it is easy to prove that (cf.[6, 31]) each interface segment/patch eK is contained in a strip of width δ and satisfies δ ≤ γ1 h2K and |ni (x) − ni (y)| ≤ γ2 hK , ∀x, y ∈ eK .

(3)

We define the weighted average {·} and the jump [·] on e ∈ Eh by {v} = κ1 v1 + κ2 v2 ,

[v] = v1 n1 + v2 n2 ,

(4)

{q} = κ1 q 1 + κ2 q 2 ,

[q] = q 1 · n1 + q 2 · n2 .

(5)

For the stability analysis of our schemes, we define (κ1 , κ2 ) on each element as follows: ⎧ |Ki | ⎪ ⎪ ⎨ 0, if |K| < c0 hK , i| 1, if |K κi = (6) |K| > 1 − c0 hK , ⎪ ⎪ |K | ⎩ i , otherwise . |K| Clearly, 0 ≤ κi ≤ 1 and κ1 + κ2 = 1 so that {·} is a convex combination along . i| Roughly speaking, we adopt the weight κi = |K |K| suggested in [13] for general sub-

elements and we set κi = 0 for |Ki | < chd+1 K . Here, the user-defined constant c0 ≥ 2γ0 γ1 and γ0 , γ1 are constants defined in (2) and (3), respectively. The dependence of c0 on these generic constants is elaborated in Lemma 1. Let χi be the characteristic function on i with i = 1, 2. Given a mesh Th , let Vh be the continuous piecewise polynomial function space of degree p ≥ 1 on the mesh. Let Vh0 := Vh ∩ H01 (), Vh1 := Vh0 · χ1 and Vh2 := Vh0 · χ2 . We define the XFE space as Vh = Vh1 + Vh2 . Then, the DG-XFE method for the interface problem is: Find uh ∈ Vh such that Bh (uh , vh ) = Fh (vh ),

∀vh ∈ Vh ,

(7)

High-Order FEMs for Interface Problems

171

where  Bh (w, v) :=

 1 ∪2

α(x)∇w · ∇v −



−β

[w] · {α(x)∇v} + 

 f v+ 

gN (κ1 v2 + κ2 v1 ) 



gD · {α(x)∇v} +

−β

 ηβ  [w] · [v], hK K∩ 

K∈Th



Fh (v) :=

{α(x)∇w} · [v] 



 ηβ  gD · [v], hK K∩ 

K∈Th

For ηβ sufficiently large, the norm corresponding to the bilinear form Bh (·, ·) is uniformly equivalent to  · Bh , which is defined by v2Bh = |v|21,1 ∪2 +

 K∈T h

2 ηβ h−1 K [v]L2 (eK ) +



ηβ−1 hK {α(x)∇v}2L2 (e ) . (8)

K∈T h

K

The crucial component in regard to establishing this equivalence result and also the stability of bilinear forms is the control on the weighted normal derivatives, which is stated as a trace and inverse inequality in Lemma 1. Lemma 1 ([27, 28]) Let γ0 and γ1 be constants defined in (2) and (3), respectively. If we choose c0 ≥ 2γ0γ1 in the definition (6) of κ, there exists a positive constant h0 such that for all h ∈ (0, h0 ] and any interface segment/patch eK = K ∩  ∈ Eh , the following estimates hold on both sub-elements of K: 1/2

κi

vi L2 (eK ) ≤

C 1/2

hK

vi L2 (Ki ) ,

vi ∈ Pp (Ki ), i = 1, 2.

(9)

The coercivity and boundedness of Bh (·, ·) in its norm  · 2Bh is then a direct consequence of the Cauchy–Schwarz inequality. Lemma 2 Let V = H 2 (1 ∪ 2 ) and V (h) = Vh + V , we have Bh (w, v) ≤ Cb wBh vBh ,

∀ w, v ∈ V (h),

(10)

and Bh (v, v) ≥ Cs v2Bh ,

∀ v ∈ Vh ,

provided the penalty parameter ηβ is chosen sufficiently large.

(11)

172

Y. Xiao et al.

The XFE space has optimal approximation quality for piecewise smooth functions in H p (1 ∪ 2 ). The following theorem is proved in [28] as an analogue of Cea’s lemma. Theorem 1 Assume that the interface  is C 2 smooth and that the solution of the elliptic interface problem (1) satisfies u ∈ H s (1 ∪ 2 ), where s ≥ 2 is an integer. Let μ = min{p + 1, s}. The following error estimates hold for any h ∈ (0, h0 ]: If ηβ is chosen sufficiently large (see (11)) and uh is the solution to the first scheme of (7), then u − uh Bh  hμ−1 uH s (1 ∪2 ) ,

∀ 0 < h ≤ h0 .

(12)

The hidden constants in the above estimates are dependent on the angle condition of the mesh Th , the degree of the polynomials, the parameter in the scheme, and α(x), but are independent of the location of the interface relative to the mesh. Here, the constant h0 is from Lemma 1.

3 An Optimal Multigrid Method for (7) In this section, we propose a two-level geometric multigrid solver of the finite element problem (7). It is well known that the element K with a “small” cut (i.e. |K ∩ i |/|K| 1) would have adverse effect on the conditioning of the resulting stiffness matrices (see e.g. [3]). Our approach is based on the general theory of the successive 0. The domain  is partitioned into grids of squares with the same size h. The exact solution is chosen as + 1/α1 exp(x1 x2 ), (x1 , x2 ) ∈ 1 , u(x1 , x2 ) = 1/α2 sin(πx1 ) sin(πx2 ), (x1 , x2 ) ∈ 2 . The right-hand side can be computed accordingly. We implement Algorithm 1, with V -cycle geometric multigrid based on the unfitted grid Th playing as the coarse grid corrector. In each pre- and post-smoothing stage of V -cycle iterator, we perform Gauss-Seidel for two times. We record the numerical results in Table 1. In these examples, the initial guess is 0, and the stopping criterion is fh − Bh uh ∞ /fh − Bh uh ∞ < 10−10. (k)

(0)

From Table 1, we can see that the multigrid method converges uniformly with respect to the mesh size, which confirms our theoretical results.

Table 1 Numerical performance of Algorithm 1 (2-D example) α1 : α2 = 1 : 10 α1 : α2 = 10 : 1

p p p p

=1 =2 =1 =2

h #iter #iter #iter #iter

2−2 7 13 24 24

2−3 10 10 30 23

2−4 10 12 29 22

2−5 11 13 27 21

2−6 12 14 25 20

High-Order FEMs for Interface Problems

175

4.3 3-D Numerical Examples The settings of this numerical experiment are as follows. The domain  = (0, 1)3 . The interfaces are two touched spheres of radius 0.1 centered at (0.4, 0.5, 0.5) and (0.6, 0.5, 0.5). The exact solution is given by ⎧ ⎨exp(x + x + x ), (x1 , x2 , x3 ) ∈ 1 , 1 2 3 u(x1 , x2 , x3 ) = ⎩sin(x1 ) sin(x2 ) sin(x3 ), (x1 , x2 , x3 ) ∈ 2 . The discontinuous coefficient function is defined such that α1 = 1 and α2 = 100. A convergence study is performed on a series of meshes generated by uniform refinements of an initial mesh consisting of 6 congruent tetrahedra. Relative errors and convergence rates of numerical solutions for Pp elements for p = 1, 2, 3 and 4 are listed in Table 2, with the quadrature order q = 2p + 3. The convergence rates are optimal for both H 1 ()-errors (order p) and L2 ()-errors (order p + 1). For the Table 2 Errors and convergence orders of the numerical solutions (3-D example) Number of Degrees of Relative H 1 error elements freedom Error Order P1 element(p = 1, q = 2p + 3 = 5) 768 189 1.690e−01 – 1241 7.510e−02 1.17 6144 49,152 9009 3.514e−02 1.10 68,705 1.658e−02 1.08 393,216 3,145,728 536,769 8.145e−03 1.03 P2 element(p = 2, q = 2p + 3 = 7) 768 1241 9.041e−03 – 9009 2.026e−03 2.16 6144 49,152 68,705 4.973e−04 2.03 536,769 1.234e−04 2.01 393,216 3,145,728 4,243,841 3.070e−05 2.01 P3 element (p = 3, q = 2p + 3 = 9) 768 3925 2.175e−03 – 6144 29,449 3.793e−05 5.84 228,241 4.743e−06 3.00 49,152 1,797,409 5.932e−07 3.00 393,216 3,145,728 14,266,945 7.414e−08 3.00 P4 element (p = 4, q = 2p + 3 = 11) 768 9009 2.971e−03 – 6144 68,705 6.042e−07 12.26 536,769 3.778e−08 4.00 49,152 393,216 4,243,841 2.362e−09 4.00

Relative L2 error Error Order 1.686e−02 3.403e−03 9.618e−04 2.272e−04 4.869e−05

– 2.31 1.82 2.08 2.22

4.150e−04 4.323e−05 5.171e−06 6.413e−07 7.965e−08

– 3.26 3.06 3.01 3.01

8.394e−05 5.864e−07 3.683e−08 2.321e−09 1.456e−10

– 7.16 3.99 3.99 3.99

9.606e−05 7.560e−09 2.380e−10 7.481e−12

– 13.63 4.99 4.99

176

Y. Xiao et al.

time being, however, the design of multigrid solver for 3-D case is still on-going. The computations for P1 , P2 and P3 elements were done using the 64-bit double precision and the linear systems were solved using MUMPS, but for P4 element, to eliminate influences of roundoff errors, the computations were done using the 80bit extended double precision and the linear systems were solved using the GMRES method with MUMPS in double precision as its preconditioner. The performance of Algorithm 1 will be reported in a future work. Acknowledgements The authors would like to acknowledge the funding support of this research by the National Key Research and Development Program of China under grant number 2017YFC0209804, National Natural Science Foundation of China under grant number 11101208, and National Center for Mathematics and Interdisciplinary Sciences of Chinese Academy of Sciences. The authors are grateful to Jinchao Xu and Haijun Wu for the fruitful discussions and suggestions, to Dr. Huaqing Liu for his valuable help on preparing the numerical examples.

References 1. Babuška, I.: The finite element method for elliptic equations with discontinuous coefficients. Computing 5, 207–213 (1970) 2. Becker, R., Burman, E., Hansbo, P.: A Nitsche extended finite element method for incompressible elasticity with discontinuous modulus of elasticity. Comput. Methods Appl. Mech. Eng. 198, 3352–3360 (2009) 3. Burman, E.: Ghost penalty. C.R. Math. 348, 1217–1220 (2010) 4. Burman, E., Hansbo, P.: Fictitious domain finite element methods using cut elements: II. A stabilized Nitsche method. Appl. Num. Math. 62, 328–341 (2012) 5. Burman, E., Guzman, J., Sanchez, M.A., Sarkis, M.: Robust flux error estimation of an unfitted Nitsche method for high-contrast interface problems. IMA J. Numer. Anal. 38, 646–668 (2018) 6. Chen, Z., Zou, J.: Finite element methods and their convergence for elliptic and parabolic interface problems. Numer. Math. 79, 175–202 (1998) 7. Chen, Z., Xiao, Y., Zhang, L.: The adaptive immersed interface finite element method for elliptic and Maxwell interface problems. J. Comput. Phys. 228, 5000–5019 (2009) 8. Chen, Z., Wu, Z., Xiao, Y.: An adaptive immersed finite element method with arbitrary Lagrangian-Eulerian scheme for parabolic equations in time variable domains. Int. J. Numer. Anal. Model. 12, 567–591 (2015) 9. Chu, C.-C., Graham, I.G., Hou, T.Y.: A new multiscale finite element method for high-contrast elliptic interface problems. Math. Comput. 79, 1915–1955 (2010) 10. Cui, T., Leng, W., Liu, H., Zhang, L., Zheng, W.: High-order numerical quadratures in a tetrahedron with an implicitly defined curved interface, ACM Transactions on Mathematical Software, in press. 11. Gong, Y., Li, B., Li, Z.: Immersed-interface finite-element methods for elliptic interface problems with non-homogeneous jump conditions. SIAM J. Numer. Anal. 46, 472–495 (2008) 12. Hansbo, P.: Nitsche’s method for interface problems in computational mechanics. GAMMMitt. 47, 183–206 (2005) 13. Hansbo, A., Hansbo, P.: An unfitted finite element method, based on Nitsche’s method for elliptic interface problems. Comput. Methods Appl. Mech. Eng. 191, 5537–5552 (2002) 14. Huang, P., Wu, H., Xiao, Y.: An unfitted interface penalty finite element method for elliptic interface problems. Comput. Methods Appl. Mech. Eng. 323, 439–460 (2017) 15. Lehrenfeld, C., Reusken, A.: L2 -error analysis of an isoparametric unfitted finite element method for elliptic interface problems (2016). Preprint. arXiv:1604.04529

High-Order FEMs for Interface Problems

177

16. Lehrenfeld, C., Reusken, A.: Analysis of a high-order unfitted finite element method for elliptic interface problems. IMA J. Numer. Anal. 38, 1351–1387 (2018) 17. LeVeque, R., Li, Z.: The immersed interface method for elliptic equations with discontinuous coefficients and singular sources. SIAM J. Numer. Anal. 31, 1019–1044 (1994) 18. Li, Z., Ito, K.: The Immersed Interface Method: Numerical Solutions of PDEs Involving Interfaces and Irregular Domains. SIAM, Philadephia (2006) 19. Li, Z., Lin, T., Wu, X.: New Cartesian grid methods for interface problems using the finite element formulation. Numer. Math. 96, 61–98 (2003) 20. Li, J., Melenk, J.M., Wohlmuth, B., Zou, J.: Optimal a priori estimates for higher order finite elements for elliptic interface problems. Appl. Numer. Math. 60, 19–37 (2010) 21. Liu, X., Fedkiw, R.P., Kang, M.: A boundary condition capturing method for Poisson’s equation on irregular domains. J. Comput. Phys. 160, 151–178 (2000) 22. Massjung, R.: An unfitted discontinuous Galerkin method applied to elliptic interface problems. SIAM J. Numer. Anal. 50(6), 3134–3162 (2012) 23. Müller, B., Kummer, F., Oberlack, M.: Highly accurate surface and volume integration on implicit domains by means of moment-fitting. Int. J. Numer. Methods Eng. 96, 512–528 (2013) 24. Peskin, C.S.: Numerical analysis of blood flow in the heart. J. Comput. Phys. 25, 220–252 (1977) 25. Saye, R.I.: High-order quadrature methods for implicitly defined surfaces and volumes in hyperrectangles. SIAM J. Sci. Comput. 37, A993–A1019 (2015) 26. The toolbox Parallel Hierarchical Grid (PHG). http://lsec.cc.ac.cn/phg 27. Wu, H., Xiao, Y.: An unfitted hp-interface penalty finite element method for elliptic interface problems. J. Comput. Math. 37, 316–339 (2019). http://arXiv.org/abs/1007.2893 28. Xiao, Y., Xu, J., Wang, F.: High-order extended finite element methods for solving interface problems. Comput. Methods Appl. Mech. Eng. 364, 112964 (2020). https://doi.org/10.1016/j. cma.2020.112964 29. Xu, J.: Estimate of the convergence rate of finite element solutions to elliptic equations of second order with discontinuous coefficients (in Chinese). Nat. Sci. J. Xiangtan Univ. 1, 1–5 (1982) 30. Xu, J.: Iterative methods by space decomposition and subspace correction. SIAM Rev. 34, 581–613 (1992) 31. Xu, J.: Estimate of the convergence rate of finite element solutions to elliptic equations of second order with discontinuous coefficients (2013, preprint). arXiv:1311.4178 32. Xu, J., Zhang, S.: Optimal finite element methods for interface problems. In: Dickopf, T., et al. (eds.), Domain Decomposition Methods in Science and Engineering XXII. Lecture Notes in Computational Science and Engineering, vol. 104, pp. 77–91. Springer, Cham (2016) 33. Xu, J., Zikatanov, L.: The method of alternating projections and the method of subspace corrections in Hilbert space. J. Am. Math. Soc. 15, 573–597 (2002)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Stabilised Hybrid Discontinuous Galerkin Methods for the Stokes Problem with Non-standard Boundary Conditions Gabriel R. Barrenechea, Michał Bosy, and Victorita Dolean

1 Introduction The interest of this paper is to discretise the Stokes problem with non-standard boundary conditions. In [1], a hybrid discontinuous Galerkin (hdG) method was proposed and analysed for this problem. The finite element method used was the combination of BDM elements of order k for the velocity, and discontinuous elements of order k − 1 for the pressure. In this paper we increase the order of the pressure space to k, while keeping the order for the velocity space fixed as k. Since this pair does not satisfy the inf-sup condition, a stabilisation term needs to be added. The stabilisation term referred to above can be built using a diversity of approaches, but, roughly speaking, the stabilisation can be residual or non-residual. In [8] the authors added a mesh-dependent term penalising the gradient of the pressure to the formulation. Later, in [14] this method was restricted and reinterpreted

G. R. Barrenechea Department of Mathematics and Statistics, University of Strathclyde, Glasgow, UK e-mail: [email protected] M. Bosy Department of Mathematics and Statistics, University of Strathclyde, Glasgow, UK Dipartimento di Matematica “F. Casorati”, Universitá degli Studi di Pavia, Pavia, Italy e-mail: [email protected] V. Dolean () Department of Mathematics and Statistics, University of Strathclyde, Glasgow, UK University Côte d’Azur, CNRS, LJAD, Nice Cedex, France e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_13

179

180

G. R. Barrenechea et al.

as a Petrov–Galerkin scheme leading to the first consistent stabilised method, and further developments were presented in the works [7] and [13]. For a review of different residual stabilised finite element methods for the Stokes problem, see the review paper [2]. Now, due to their nature, residual methods include unphysical couplings to the formulation, and modify all the entries of the stiffness matrix. Hence, non-residual methods where only a positive semi-definite term penalising the pressure is added have also being proposed. Examples of this type of methods are the pressure gradient projection [9] and local pressure gradient stabilisation [3]. The methods just mentioned typically use two nested meshes in order to build the method. Thus, to avoid this complication, the local pressure gradient stabilisation has been also presented on the same mesh in [12]. Additionally, methods that use fluctuations of the pressure gradient are not effective when the finite element space for pressure is the piecewise constant space. The usual way to overcome this is to add pressure jumps to the formulation, as it has been done, e.g., in [16]. These have been shown to be very effective, but they do somehow temper with the data structure of the code. To avoid this, the authors in [10] present an approach that is based on polynomialpressure-projection. This method works for low order of polynomials as was shown in [4], and preserves symmetry of the original equation. In the light of the discussion of the previous paragraphs, in this work we propose a stabilised hdG method for the Stokes problem with non-standard boundary conditions. The method is reminiscent of the Dorhmann–Bochev method (from [10]), but uses the same velocity space used in the hdG method from [1].

1.1 Notations and Model Problem Let  be an open polygonal domain in R2 with Lipschitz boundary  := ∂. We use boldface font for tensor or vector variables e.g. u is a velocity vector field. The scalar variables will be italic e.g. p denotes pressure scalar value. We define the stress tensor σ := ν∇u − pI (where ν > 0 is the fluid viscosity and I is the identity matrix) and the flux as σn := σ n. In addition, we denote normal and tangential components as follows un := u · n, ut := u · t, σnn := σn · n, where n is the outward unit normal vector to the boundary  and t is a vector tangential to  such that n · t = 0. For D ⊂ , we use the standard L2 (D) space with the following norm  f 2D :=

f 2 dx for all f ∈ L2 (D). D

Stabilised HDG Methods for the Stokes Problem with Non-standard B. C.

181

Let us define, for m ∈ N, the following Sobolev spaces   H m (D) := v ∈ L2 (D) : ∀ |α| ≤ m ∂ α v ∈ L2 (D) ,   H (div, D) := v ∈ [L2 (D)]2 : ∇ · v ∈ L2 (D) , ∂ |α| α α . In addition, ∂x1 1 ∂x2 2 Sobolev space H m (D)

where, for α = (α1 , α2 ) ∈ N2 , |α| = α1 + α2 , and ∂ α = will use the standard semi-norm and norm for the |f |2H m (D) :=



∂ α f 2D ,

f 2H m (D) :=

|α|=m

m 

we

|f |2H k (D) ∀ f ∈ H m (D).

k=0

In this work, we consider the two dimensional Stokes problem with tangentialvelocity and normal-flux (TVNF) boundary conditions ⎧ ⎪ ⎪ −ν u + ∇p ⎪ ⎨ ∇ ·u ⎪ σnn ⎪ ⎪ ⎩ ut

=f =0 =g =0

in , in , on , on ,

(1)

¯ → R2 is the unknown velocity field, p :  ¯ → R the pressure, ν > 0 where u :  the viscosity, which is considered to be constant, and f ∈ [L2 ()]2, g ∈ L2 () are given functions. The restriction to homogeneous Dirichlet conditions on ut is made only to =simplify the presentation. > ¯ made of triangles. For Let Th h>0 be a regular family of triangulations of  each triangulation Th , Eh denotes the set of its edges. In addition, for each of element K ∈ Th , hK := diam(K), and we denote h := maxK∈Th hK . We define following Sobolev spaces on the triangulation Th and the set of all edges in Eh   L2 (Eh ) := v : v|E ∈ L2 (E) ∀ E ∈ Eh ,   H m (Th ) := v ∈ L2 () : v|K ∈ H m (K) ∀ K ∈ Th for m ∈ N, with the corresponding broken norms. Now we will introduce the finite element spaces that discretise the above spaces. Let k ≥ 1. We start by introducing the velocity and pressure spaces. To discretise the velocity u we use the Brezzi–Douglas–Marini space (see [5, Section 2.3.1]) of order k ≥ 1 defined by    2 BDMhk := vh ∈ H (div, ) : vh |K ∈ Pk (K) ∀ K ∈ Th .

182

G. R. Barrenechea et al.

Associated to this space, we introduce the BDM projection +k : [H 1 ()]2 → BDMhk defined in [5, Section 2.5]. The pressure is discretised using the following space   Qkh := qh ∈ L2 () : qh |K ∈ Pk (K) ∀ K ∈ Th . k Associated to this space we define the local L2 (K)-projection K : L2 (K) → k 2 Pk (K) for each K ∈ Th defined as follows. For every w ∈ L (K), K (w) is ; ; k the unique element of Pk (K) satisfying K K (w)vh dx = K wvh dx ∀ vh ∈ k for all K ∈ T . Pk (K) , and we define the continuous projection  k |K = K h The last ingredient needed in the method described below is a finite element space associated to a family of Lagrange multipliers associated to the edges of the triangulation. These multipliers will be denoted by u˜ and are meant to approximate the tangential trace of the velocity u on the edges of the triangulation. For this, and in order to propose a discretisation with fewer degrees of freedom, we discretise the Lagrange multiplier u˜ using the space

    k−1 Mh,0 := v˜h ∈ L2 Eh : v˜h |E ∈ Pk−1 (E) ∀ E ∈ Eh , v˜h = 0 on  . Furthermore, we introduce for all E ∈ Eh the L2 (E)-projection k−1 : L2 (E) → E k−1 2 ˜ is the unique element Pk−1 (E) defined as follows. For every w˜ ∈ L (E), E (w) ; ; ( w) ˜ v ˜ ds = w ˜ v ˜ ds ∀ v˜h ∈ Pk−1 (E) , and we of Pk−1 (E) satisfying E k−1 h h E E   k−1 k−1 k−1 2 k−1 : L Eh → Mh defined as  |E := E for all E ∈ Eh . denote 

2 The Stabilised Method Our approach is to write the discrete problem with the same degree of polynomials k−1 for velocity and pressure spaces. In other words, denoting Vh := BDMhk × Mh,0 , k−1 k we want to use the space Vh × Qh , instead of Vh × Qh as it was done in [1]. To do this, we need the proper stabilisation term, because this choice of spaces does not guarantee inf-sup stability. The first ingredient in the definition of the stabilised method for (1) we use the same bilinear forms as in [1], this is       a wh , w˜ h , vh , v˜h := ν∇wh : ∇vh dx  − ∂K

K∈Th

K

    ν ∂n wh t (vh )t − v˜h ds + ε

 ∂K

   ν (wh )t − w˜ h ∂n vh t ds

Stabilised HDG Methods for the Stokes Problem with Non-standard B. C.

183

   k−1   τ k−1 +ν  (wh )t − w˜ h  (vh )t − v˜h ds hK ∂K      qh ∇ · vh dx, b vh , v˜h , qh := − K∈Th K

where ε ∈ {−1, 1} and τ > 0 is a stabilisation parameter. In addition, to compensate for the non-inf-sup stability of the finite element spaces we have chosen, we introduce the bilinear form       1 s ph , qh := ph −  k−1 ph qh −  k−1 qh dx. ν  With these ingredients we can method   now present the finite element  analysed in  this work: Find uh , u˜ h , ph ∈ Vh × Qkh such that for all vh , v˜h , qh ∈ Vh × Qkh A



     uh , u˜ h , ph , vh , v˜h , qh = f vh dx + g(vh )n ds, 

(2)



where            A uh , u˜ h , ph , vh , v˜h , qh :=a uh , u˜ h , vh , v˜h + b vh , v˜h , ph      + b uh , u˜ h , qh − s ph , qh .

2.1 Well-Posedness of the Discrete Problem Let us consider the following norm on Vh (see [1, Lemma 3.2] for a proof that this is actually a norm in Vh )

  2    τ 2 k−1  |wh |2H 1 (K) + hK ∂n wh ∂K + ||| wh , w˜ h |||2 :=ν (wh )t − w˜ h .  ∂K hK K∈Th

The first step towards proving the stability of Method (2) is the following weak inf-sup condition for b. Lemma 1 There exist constants C1 , C2 > 0, independent of hK and ν, such that b sup (vh ,v˜h )∈Vh

  vh , v˜h , qh   ≥ C1 qh  − C2 qh −  k−1 qh  ||| vh , v˜h ||| 

∀qh ∈ Qkh . (3)

184

G. R. Barrenechea et al.

˜ be a convex, open, Lipschitz set Proof We consider an arbitrary qh ∈ Qkh . Let  ˜ such that  ⊂ , and let us consider following extension + qh in  qˆh := ˜ \ . 0 in  Let now φ be the unique weak solution of the problem + ˜ − φ = qˆh in  . φ = 0 on ∂ ˜ is convex, then φ ∈ H 2 (). ˜ Then w := ∇φ| belongs to [H 1 ()]2, and Since  for w˜ := wt , b

   w, w˜ , qh = qh 2

∀qh ∈ Qkh .

(4)

In addition, applying standard regularity results, see [5, Section 1.2], we get wH 1 () ≤ ∇φH 1 () ˜ ≤ c1 qh  .

(5)

"2 ! In [1, Lemma 3.5] it is shown that there exists a Fortin operator : H 1 () → Vh satisfying the following condition: for all v ∈ [H 1 ()]2 the following holds b

     v, v˜ , qh = b (v) , qh ∀ qh ∈ Qk−1 h , √ ||| (v) ||| ≤ C νvH 1 () .

(6) (7)

  Let wh , w˜ h := (w), then thanks to (6), (4) and the continuity of b (see [1, Lemma 3.3])          b wh , w˜ h , qh = b w, w˜ , qh − b w − wh , w˜ − w˜ h , qh −  k−1 qh  |wh − w|2H 1 (K) qh −  k−1 qh . ≥ qh 2 − c2 

K∈Th

Using the approximation properties of the BDM interpolation operator (see [5, Preposition 2.5.1]) and (5)  1   k−1 qh  − c2 c3 qh −  qh |w|H 1 () b wh , w˜ h , qh ≥  c1    ≥ C1 qh  − C2 qh −  k−1 qh ||| wh , w˜ h ||| , 

Stabilised HDG Methods for the Stokes Problem with Non-standard B. C.

185

where, in the last estimate we have used the stability of the Fortin operator in the ||| · ||| norm (7). This proves the result with C1 = C √1νc and C2 = Cc2√c3ν . & % 1

Before showing an inf-sup condition, we prove the continuity of bilinear form A.     Lemma 2 There exists a constant C > 0 such that, for all wh , w˜ h , vh , v˜h ∈ Vh and rh , qh ∈ Qkh , we have             A wh , w˜ h , rh , vh , v˜h , qh  ≤ C||| wh , w˜ h , rh |||h ||| vh , v˜h , qh |||h .   (8) Proof We use the continuity of the bilinear forms (see [1, Lemma 3.3]) and the fact that the projection is a bounded operator. & % The final step towards stability is proving the inf-sup condition for bilinear form A.   Lemma 3 There exists β > 0 independent of hK such that for all wh , w˜ h , rh ∈ Vh × Qkh the following holds A sup (vh ,v˜h ,qh )∈Vh ×Qkh

    wh , w˜ h , rh , vh , v˜h , qh     ≥ β||| wh , w˜ h , rh |||h . ||| vh , v˜h , qh |||h

(9)

As a consequence, Problem (2) is well-posed.   Proof Let wh , w˜ h , rh ∈ Vh × Qkh . The idea of the proof is to construct an appropriate vh , v˜h , qh such that A

        wh , w˜ h , rh , vh , v˜h , qh ≥ c||| wh , w˜ h , rh |||h ||| vh , v˜h , qh |||h .

To achieve that we use coercivity of a (see [1, Lemma 3.4]), continuity of a (see [1, Lemma 3.3]) and Lemma 2. For details see [6]. & %

2.2 Error Analysis In this section we present the error estimates for the method. The addition of the stabilising bilinear form s(·, ·) introduced a consistency error. However according to [4], this should not be viewed as a serious flaw, as this consistency error can be bounded in an optimal way. The following result is the first step towards that goal. !  "2   × L2 () be the solution of the Lemma 4 Let u, p ∈ H 1 () ∩ H 2 Th   problem (1) and u˜ = ut on all edges of Eh . If uh , u˜ h , ph ∈ Vh × Qkh solves (2),

186

G. R. Barrenechea et al.

  then for all vh , v˜h , qh ∈ Vh × Qkh the following holds A

      u − uh , u˜ − u˜ h , p − ph , vh , v˜h , qh = s p, qh .

(10)

Next, we introduce the following norm 1 |||(u, u, ˜ p)|||h := |||(u, u)||| ˜ + √ p , ν

(11)

and prove the following variant of Cea’s lemma [11, Lemma 2.28] for this stabilised Stokes problem. !    "2 Lemma 5 Let u, p ∈ H 1 () ∩ H 2 Th × L2 () be the solution of the   problem (1) and u˜ = ut on all edges of Eh . If uh , u˜ h , ph ∈ Vh × Qkh solves (2), then there exists C > 0, independent of h and ν, such that   ||| u − uh , u˜ − u˜ h , p − ph |||h ≤C

  inf ||| u − vh , u˜ − v˜h , p − qh |||h k (vh ,v˜h ,qh )∈Vh ×Qh C + √ p −  k−1 p . (12)  ν

Proof It is a combination of Lemmas 1, 2 and 3. For details see [6]. & % ! "     2 Lemma 6 Let u, p ∈ H 1 () ∩ H 2 Th × L2 () be the solution of the   problem (1) and u˜ = ut on all edges of Eh . If uh , u˜ h , ph ∈ Vh × Qkh solves (2), then there exists C > 0, independent of h and ν, such that # $   1 k √ ||| u − uh , u˜ − u˜ h , p − ph |||h ≤ Ch νuH k+1 (Th ) + √ pH k (Th ) . ν Proof It is a combination of [1, Lemmas 3.8] and Lemma 5 with the local L2 projection approximation [11, Theorem 1.103]. % &

3 Numerical Experiments The computational domain is the unit square  = (0, 1)2 . We present the results 0 for k = 1, that is the discrete space is given by BDMh1 × Mh,0 × Q1h . We test both the symmetric method (ε = −1) and the non-symmetric method (ε = 1). We have followed the recommendation given in [15, Section 2.5.2] and taken τ = 6.

Stabilised HDG Methods for the Stokes Problem with Non-standard B. C.

187

We choose the right hand side f and the boundary condition g such that the exact solution is given by u = curl



   1 − cos((1 − x)2 ) sin(x 2 ) sin(y 2 ) 1 − cos((1 − y)2 ) ,

p = tan(xy).

In Fig. 1a and b we depict the errors for both the symmetric and non-symmetric cases, respectively. We can see that they not only validate the theory from Sect. 2.2, but also perform an optimal h2 convergence rate for u − uh  . Furthermore, we observe an increased order of convergence for p − ph  . In fact, the error seems to decrease with O(h3/2 ), rather than the O(h) predicted by the theory. To stress the last point made in the previous paragraph, in Table 1 we compare the L2 error of the pressure (||p − ph || ) for hdG method introduced in [1] and stabilised hdG method from Sect. 2. Columns ph ∈ Q0h are associated with hdG method and ph ∈ Q1h with stabilised hdG ones. There, we confirm that the pressure

1

0

0

-1

log(error)

log(error)

-1 -2 -3 -4

-2 -3 -4

-5

-5

-6

-6

-2.5

-2

-1.5

-1

-0.5

-2.5

-2

-1.5

-1

-0.5

log(h)

log(h)

(b)

(a)

Fig. 1 Convergence the stabilised method with k = 1. (a) Symmetric bilinear form (ε = −1). (b) Non-symmetric bilinear form (ε = 1) Table 1 Comparison of the error of the pressure ||p − ph || h 2−1 2−2 2−3 2−4 2−5 2−6 2−7 2−8

Symmetric bilinear form (ε = −1) ph ∈ Q0h ph ∈ Q1h 0.152296 0.077228 0.082775 0.041790 0.042620 0.020500 0.021357 0.008338 0.010676 0.003083 0.005340 0.001105 0.002671 0.000392 0.001336 0.000139

Non-symmetric bilinear form (ε = 1) ph ∈ Q0h ph ∈ Q1h 0.159019 0.090624 0.084875 0.047488 0.043313 0.009449 0.021513 0.003516 0.010707 0.001269 0.005346 0.002171 0.002672 0.000453 0.001336 0.000161

188

G. R. Barrenechea et al.

error for the stabilised version is much smaller than the one for the inf-sup stable case, in addition to having an increased order of convergence.

4 Conclusion In this work we have applied the idea introduced in [10] to stabilise the hdG method proposed in [1] for the Stokes problem with TVNF boundary conditions. The method adds a simple, symmetric, term to the formulation, and allowed us to use a higher order pressure space, which, in turn, improved the pressure convergence (although a proof of this fact is, in general, not available). This approach was also applied to NVTF boundary conditions (see [6]) and can be used for other discontinuous Galerkin methods that deal with Stokes or nearly incompressible elasticity problems. Future testing using higher order discretisations is needed to assess whether this approach provides an increase of the convergence rate for the pressure. Thus, the numerical tests with higher order of polynomials for discontinuous finite methods is interest for further research to look for the improvement of the convergence.

References 1. Barrenechea, G.R., Bosy, M., Dolean, V., Nataf, F., Tournier, P.-H.: Hybrid discontinuous Galerkin discretisation and domain decomposition preconditioners for the Stokes problem. Comput. Methods Appl. Math. 19(4), 703–722 (2019) 2. Barth, T., Bochev, P.B., Gunzburger, M., Shadid, J.: A taxonomy of consistently stabilized finite element methods for the Stokes problem. SIAM J. Sci. Comput. 25(5), 1585–1607 (2004) 3. Becker, R., Braack, M.: A finite element pressure gradient stabilization for the Stokes equations based on local projections. Calcolo 38(4), 173–199 (2001) 4. Bochev, P.B., Dohrmann, C.R., Gunzburger, M.D.: Stabilization of low-order mixed finite elements for the Stokes equations. SIAM J. Numer. Anal. 44(1), 82–101 (2006) 5. Boffi, D., Brezzi, F., Fortin, M.: Mixed Finite Element Methods and Applications. Springer Series in Computational Mathematics, vol. 44. Springer, Heidelberg (2013) 6. Bosy, M.: Efficient discretisation and domain decomposition preconditioners for incompressible fluid mechanics. Ph.D. Thesis, University of Strathclyde (2017). https://doi.org/10.13140/ RG.2.2.24947.17444 7. Brezzi, F., Douglas, J., Jr.: Stabilized mixed methods for the Stokes problem. Numer. Math. 53(1–2), 225–235 (1988) 8. Brezzi, F., Pitkäranta, J.: On the stabilization of finite element approximations of the Stokes equations. In: Efficient Solutions of Elliptic Systems (Kiel, 1984). Notes on Numerical Fluid Mechanics, vol. 10, pp. 11–19. Friedr. Vieweg, Braunschweig (1984) 9. Codina, R., Blasco, J.: Analysis of a pressure-stabilized finite element approximation of the stationary Navier-Stokes equations. Numer. Math. 87(1), 59–81 (2000) 10. Dohrmann, C.R., Bochev, P.B.: A stabilized finite element method for the Stokes problem based on polynomial pressure projections. Int. J. Numer. Methods Fluids 46(2), 183–201 (2004) 11. Ern, A., Guermond, J.L.: Theory and Practice of Finite Elements. Applied Mathematical Sciences, vol. 159. Springer, New York (2004)

Stabilised HDG Methods for the Stokes Problem with Non-standard B. C.

189

12. Ganesan, S., Matthies, G., Tobiska, L.: Local projection stabilization of equal order interpolation applied to the Stokes problem. Math. Comput. 77(264), 2039–2060 (2008) 13. Hughes, T.J.R., Franca, L.P.: A new finite element formulation for computational fluid dynamics. VII. The Stokes problem with various well-posed boundary conditions: symmetric formulations that converge for all velocity/pressure spaces. Comput. Methods Appl. Mech. Eng. 65(1), 85–96 (1987) 14. Hughes, T.J.R., Franca, L.P., Balestra, M.: A new finite element formulation for computational fluid dynamics. V. Circumventing the Babuška-Brezzi condition: a stable Petrov-Galerkin formulation of the Stokes problem accommodating equal-order interpolations. Comput. Methods Appl. Mech. Eng. 59(1), 85–99 (1986) 15. Lehrenfeld, C.: Hybrid discontinuous Galerkin methods for solving incompressible flow problems. Dissertation, Rheinisch-Westfälischen Technischen Hochschule Aachen (2010) 16. Silvester, D.J., Kechkar, N.: Stabilised bilinear-constant velocity-pressure finite elements for the conjugate gradient solution of the Stokes problem. Comput. Methods Appl. Mech. Eng. 79(1), 71–86 (1990)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

RBF Based CWENO Method Jan S. Hesthaven, Fabian Mönkeberg, and Sara Zaninelli

1 Introduction A broad range of physical phenomena can be described by hyperbolic conservation laws of the form ut + f (u)x = 0,

(x, t) ∈ R × R+ ,

u(0) = u0 ,

(1)

with the conserved variables u : R × R+ → RN and the flux function f : RN → RN . The nonlinear behavior of f can lead to complex solutions, most notably shocks. It is well-known that high-order methods give good results for smooth data, but for discontinuous ones spurious oscillations are introduced. A popular class of methods to solve (1) is the finite volume method, which is based on a discretization in space . . . < xi−1/2 < xi+1/2 < . . . and the average values u¯ i of its cells Ci = [xi−1/2 , xi+1/2 ]. It is defined by the semi-discrete scheme Fi+1/2 − Fi−1/2 du¯ i =− , dt Δx

(2)

where the numerical flux term Fi+1/2 depends on the values {u¯ i−k , . . . , u¯ i+p−k } with 0 ≤ k ≤ p − 1. For more details we refer the reader to [15, 20, 22]. The class of essentially nonoscillatory (ENO) methods, introduced by Harten et al. [14], reduces spurious oscillations to a minimum. They are based on a monotone numerical flux function F (u, v) and high-order accurate reconstruction si (x) for J. S. Hesthaven · F. Mönkeberg () · S. Zaninelli École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland e-mail: [email protected]; [email protected]; [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_14

191

192

J. S. Hesthaven et al.

each cell i. The central idea is to choose the least oscillating interpolation function − ± si and define the numerical flux Fi+1/2 = F (u+ i+1/2 , ui+1/2 ) with ui+1/2 being the evaluation of si+1 and si at the interface xi+1/2 . Based on the ENO method, Jiang and Shu [19] introduced the weighted ENO (WENO) method which considers different interpolation polynomials, based on different stencils, and combines them in a nonoscillatory manner to maximize the attainable accuracy. Further results on ENO and WENO methods can be found in [10, 11, 16].

2 CWENO The CWENO method is based on the WENO method and was introduced by Levy et al. [23] as a third order method. Further analysis and generalization to higher orders on general grids can be found in [6, 7]. Let us consider the standard semi-discrete formulation (2) with a monotone flux function F (u, v). The goal is to construct a reconstruction Prec,i for each cell Ci based on the stencil {Ci−k , . . . , Ci+k } for k ∈ N. In the smooth regions the algorithm should choose a polynomial of degree 2k which interpolates the central stencil u¯ i−k , . . . , u¯ i+k in the mean value sense. In case of a non-smooth solution it chooses a polynomial of degree k on one stencil {Ci−k+l , . . . , Ci+l } that avoids the discontinuity. Given the reconstruction, the high-order numerical flux is Fi+1/2 = F (Prec,i+1 (xi+1/2 ), Prec,i (xi+1/2 )). Specifically, let us consider Popt as the polynomial of degree 2k that interpolates all data in the 2k + 1 stencil and the polynomials Pl of degree k that interpolate the data on the stencil {Ci−k+l−1 , . . . , Ci+l−1 } for l = 1, . . . , k + 1. Furthermore, the reconstruction depends on the choice of the positive real coefficients d0 , . . . , dk+1 ∈ < [0, 1] such that k+1 d = 1, d0 = 0. Then, the reconstruction polynomial of degree l l=0 2k is Prec (x) =

k+1 

ωl Pl (x),

(3)

l=0

with   1 Popt (x) − dl Pl (x) , d0 k+1

P0 (x) =

(4)

l=1

and the nonlinear coefficients ωl that are defined as αl ωl = 0. For #¯ = #h ˆ 2 with #ˆ = 0.1 we get the right order of convergence, but for the 7th order method (k = 3) we choose #ˆ = 10−6 to reduce spurious oscillations for the Euler equations. Moreover, we should point out that the choice of the linear weight d0 can influence the result; indeed if it is too close to 1 then the reconstruction almost coincides with Popt , which can lead to spurious oscillations in case of discontinuous solutions. We present multiple numerical examples to show the robustness of the method. We can conclude that the RBF-CWENO method works comparable to the existing RBF-WENO and ENO methods in one dimension. The advantage of RBFs is clearer when considering unstructured grids in higher dimensions where polynomial reconstruction is complex.

References 1. Aboiyar, T., Georgoulis, E.H., Iske, A.: High order WENO finite volume schemes using polyharmonic spline reconstruction. In: Proceedings of the International Conference on Numerical Analysis and Approximation Theory NAAT2006, Cluj-Napoca (Romania). Department of Mathematics, University of Leicester (2006) 2. Aboiyar, T., Georgoulis, E.H., Iske, A.: Adaptive ADER methods using kernel-based polyharmonic spline WENO reconstruction. SIAM J. Sci. Comput. 32(6), 3251–3277 (2010) 3. Bigoni, C., Hesthaven, J.S.: Adaptive WENO methods based on radial basis function reconstruction. J. Sci. Comput. 72(3), 986–1020 (2017) 4. Buhmann, M.D.: Radial Basis Functions: Theory and Implementations. Cambridge University Press, Cambridge (2003) ˇ 5. Chandhini, G., Sanyasiraju, Y.: Local RBF-RFD solutions for steady convection–diffusion problems. Int. J. Numer. Methods Eng. 72(3), 352–378 (2007) 6. Cravero, I., Semplice, M.: On the accuracy of WENO and CWENO reconstructions of third order on nonuniform meshes. J. Sci. Comput. 67(3), 1219–1246 (2016) 7. Cravero, I., Puppo, G., Semplice, M., Visconti, G.: CWENO: uniformly accurate reconstructions for balance laws (2016). Preprint. arXiv:1607.07319 8. Driscoll, T.A., Fornberg, B.: Interpolation in the limit of increasingly flat radial basis functions. Comput. Math. Appl. 43(3), 413–422 (2002) 9. Duchon, J.: Splines Minimizing Rotation-Invariant Semi-Norms in Sobolev Spaces, pp. 85– 100. Springer, Berlin (1977) 10. Fjordholm, U.S., Ray, D.: A sign preserving WENO reconstruction method. J. Sci. Comput. 1–22 (2016) 11. Fjordholm, U.S., Mishra, S., Tadmor, E.: Arbitrarily high-order accurate entropy stable essentially nonoscillatory schemes for systems of conservation laws. SIAM J. Numer. Anal. 50(2), 544–573 (2012) 12. Fornberg, B., Lehto, E., Powell, C.: Stable calculation of gaussian-based RBF-FD stencils. Comput. Math. Appl. 65(4), 627–637 (2013) 13. Hardy, R.L.: Multiquadric equations of topography and other irregular surfaces. J. Geophys. Res. 76(8), 1905–1915 (1971) 14. Harten, A., Engquist, B., Osher, S., Chakravarthy, S.R.: Uniformly high order accurate essentially non-oscillatory schemes, iii. J. Comput. Phys. 71(2), 231–303 (1987) 15. Hesthaven, J.S.: Numerical Methods for Conservation Laws: From Analysis to Algorithms. Society for Industrial and Applied Mathematics, Philadelphia (2017)

RBF Based CWENO Method

201

16. Hu, C., Shu, C.-W.: Weighted essentially non-oscillatory schemes on triangular meshes. J. Comput. Phys. 150(1), 97–127 (1999) 17. Iske, A.: Multiresolution Methods in Scattered Data Modelling, vol. 37. Springer, Berlin (2004) 18. Iske, A., Sonar, T.: On the structure of function spaces in optimal recovery of point functionals for ENO-schemes by radial basis functions. Numer. Math. 74(2), 177–201 (1996) 19. Jiang, G.-S., Shu, C.-W.: Efficient implementation of weighted ENO schemes. J. Comput. Phys. 126(1), 202–228 (1996) 20. Kröner, D.: Numerical Schemes for Conservation Laws. Wiley, Chichester (1997) 21. Larsson, E., Fornberg, B.: Theoretical and computational aspects of multivariate interpolation with increasingly flat radial basis functions. Comput. Math. Appl. 49(1), 103–130 (2005) 22. LeVeque, R.J.: Numerical Methods for Conservation Laws. Springer Science & Business Media, New York (1992) 23. Levy, D., Puppo, G., Russo, G.: Central WENO schemes for hyperbolic systems of conservation laws. ESAIM: Math. Model. Numer. Anal. 33(3), 547–571 (1999) 24. Micchelli, C.A.: Interpolation of scattered data: distance matrices and conditionally positive definite functions. Constr. Approx. 2(1), 11–22 (1986) 25. Mönkeberg, F., Hesthaven, J.S.: Entropy stable essentially nonoscillatory methods based on RBF reconstruction. ESAIM: Math. Model. Numer. Anal. 53, 925–958 (2019) 26. Schaback, R.: Multivariate interpolation by polynomials and radial basis functions. Constr. Approx. 21(3), 293–317 (2005) 27. Wendland, H.: Scattered Data Approximation. Cambridge University Press, Cambridge (2004) 28. Wright, G.B., Fornberg, B.: Stable computations with flat radial basis functions using vectorvalued rational approximations. J. Comput. Phys. 331, 137–156 (2017)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Discrete Equivalence of Adjoint Neumann–Dirichlet div-grad and grad-div Equations in Curvilinear 3D Domains Yi Zhang, Varun Jain, Artur Palha, and Marc Gerritsma

1 Introduction In Rd , given a bounded domain  with Lipschitz boundary ∂ and σˆ n ∈ H −1/2(∂) = tr H (div, ), ω ∈ H 1 () solves the Neumann problem, ⎧ ⎪ ⎨

∂ω = σˆ n ∂n

⎪ ⎩ −div grad ω + ω = 0

on ∂

,

(1)

in 

if and only if σ ∈ H (div, ) which solves the Dirichlet problem, ⎧ ⎨

σ · n = σˆ n

⎩ −grad (div σ ) + σ = 0

on ∂ in 

,

(2)

satisfies σ = grad ω [3]. This is obvious at the continuous level. The question is whether we can find a set of finite dimensional function spaces such that σ h = grad ωh holds if ωh and σ h solve the discrete Neumann and Dirichlet problems respectively. The answer is yes.

Y. Zhang () · V. Jain · M. Gerritsma Delft University of Technology, Delft, Netherlands e-mail: [email protected]; [email protected]; [email protected] A. Palha Eindhoven University of Technology, Eindhoven, Netherlands e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_15

203

204

Y. Zhang et al.

Throughout this paper, we restrict ourselves to R3 . We will first construct the primal polynomial spaces and their algebraic dual representations, and then use them to discretize problems (1) and (2) such that the identity σ h = grad ωh holds at the discrete level in any curvilinear domain for any polynomial approximation degree. This work extends [7, 9], where similar dual Neumann–Dirichlet problems are considered, to 3-dimensional space. These primal spaces and their algebraic dual representations can be ideal for the so-called mimetic or structure-preserving discretizations [1, 4, 8, 11, 12]. Together with their trace spaces, they can be used for the hybrid finite element methods which first decompose the domains into discontinuous elements then connect them with Lagrange multipliers living in the trace spaces [2, 13, 14]. The outline of this paper is as follows: In Sect. 2, we introduce the construction of polynomial spaces and their algebraic dual representations. The discrete formulations of the Neumann–Dirichlet problems and the proof of their equivalence at the discrete level follow in Sect. 3. A 3-dimensional numerical test case is then presented in Sect. 4. Finally, conclusions are drawn in Sect. 5.

2 Function Spaces 2.1 Primal Polynomial Spaces Let −1 = ξ0i < ξ1i < · · · < ξIi i = 1, i = 1, 2, 3, being three partitionings of [−1, 1]. The associated Lagrange polynomials are i

hj (ξ ) = i

I @

ξ i − ξmi

m=0,m=j

ξji − ξmi

, j = 0, 1, · · · , I i .

They are polynomials of degree I i which satisfy the Kronecker delta property, hj (ξki ) = δj k . The associated edge functions can be derived as [6], j −1  dhk (ξ i )

ej (ξ ) = − i

k=0

dξ i

, j = 1, 2, · · · , I i ,

which are polynomials of degree I i − 1. Edge functions also satisfy the Kronecker delta property, but in the integral sense, 

ξki

i ξk−1

ej (ξ i ) dξ i = δj k .

Discrete Equivalence of Adjoint Neumann–Dirichlet Equations

205

Consider a reference domain ref |ξ 1 ,ξ 2 ,ξ 3 := [−1, 1]3. With the tensor product, we can construct finite dimensional scalar function space PI polynomial basis functions 

1 ,I 2 ,I 3

spanned by

 hi (ξ 1 )hj (ξ 2 )hk (ξ 3 ) ,

and vector-valued function space LI

1 ,I 2 ,I 3

spanned by polynomial basis functions

  ei (ξ 1 )hj (ξ 2 )hk (ξ 3 ), hi (ξ 1 )ej (ξ 2 )hk (ξ 3 ), hi (ξ 1 )hj (ξ 2 )ek (ξ 3 ) . Let ωh ∈ PI

1 ,I 2 ,I 3

be 1

ω = h

2

3

I  I I  

wi,j,k hi (ξ 1 )hj (ξ 2 )hk (ξ 3 ).

(3)

i=0 j =0 k=0

Due to the way of constructing the edge functions, we can easy derive ρ h = 1 2 3 grad ωh ∈ LI ,I ,I , ρ h = grad ωh = (ρ1 , ρ2 , ρ3 )T , where [6], 1

2

3

I  I  I    ρ1 = wi,j,k − wi−1,j,k ei (ξ 1 )hj (ξ 2 )hk (ξ 3 ), i=1 j =0 k=0 1

ρ2 =

2

3

I  I I     wi,j,k − wi,j −1,k hi (ξ 1 )ej (ξ 2 )hk (ξ 3 ), i=0 j =1 k=0 1

2

3

I  I  I    wi,j,k − wi,j,k−1 hi (ξ 1 )hj (ξ 2 )ek (ξ 3 ). ρ3 = i=0 j =0 k=1

Let ω, ρ be the vectors of expansion coefficients of ωh , ρ h . We can obtain ρ = E ω,

(4)

where E is called the incidence matrix. The incidence matrix is very sparse, only consists of ±1 as non-zero entries. If we squeeze, stretch or distort the domain, of course, the polynomial basis functions change, but the incidence matrix will remain the same. It only depends on the topology of the mesh and the numbering of the

206

Y. Zhang et al.

degrees of freedom. And it is exact. In other words, it introduces no extra error. All these features make it an excellent discrete counterpart of the grad operator. Examples of incidence matrices can be found in [8, 10–12]. For a comprehensive explanation of these polynomial basis functions, we refer to [6]. In isogeometric analysis, tensor-product B-splines with similar properties have been developed, see, for example [5]. For tetrahedral elements, an analogue development can be found in [15]. h From (3),  we can derive the trace of ω  , for example, on the back boundary of ref , b = ξ 1 = −1, ξ 2 , ξ 3 ∈ [−1, 1] , 2

trb ωh =

3

I  I 

w0,j,k h0 (−1)hj (ξ 2 )hk (ξ 3 ).

j =0 k=0

Let ωb be the vector of expansion coefficients of trb ωh . Clearly, there exists a linear operator Nb such that ωb = Nb ω. The same processes can be done for other boundaries. If we collect the traces of ωh on all boundaries and combine their vectors of expansion coefficients and corresponding linear operators, we can eventually obtain ωtr = N ω, where the matrix N, like E, is sparse and only depends on the topology of the mesh and the numbering of the degrees of freedom. Furthermore, it contains only 1 as non-zero entries. An example of N can be found in [7]. Now, we can conclude that 1 2 3 1 2 3 the trace space, PI ,I ,I = tr PI ,I ,I , is given as PI

1 ,I 2 ,I 3

2

3

:= PI−1,I ∪ PI1

2 ,I 3

1

3

∪ PI−1,I ∪ PI1

1 ,I 3

1

2

∪ PI−1,I ∪ PI1

1 ,I 2

,

  2 3 is the space spanned by h0 (−1)hj (ξ 2 )hk (ξ 3 ) , PI1 ,I is the   space spanned by hI 1 (1)hj (ξ 2 )hk (ξ 3 ) and so on. Notice that the polynomial   basis functions in h0 (−1)hj (ξ 2 )hk (ξ 3 ) are exactly the same as those in   hI 1 (1)hj (ξ 2 )hk (ξ 3 ) because h0 (−1) = hI 1 (1) = 1. But here we still distinguish them because they represent basis functions at different boundaries. 2

where PI−1,I

3

Discrete Equivalence of Adjoint Neumann–Dirichlet Equations

207

2.2 Algebraic Dual Polynomial Spaces We first consider the space PI example,

1 ,I 2 ,I 3

. Let MP be the symmetric mass matrix, for

MP i+j (I 1 +1)+k(I 1+1)(I 2 +1), l+m(I 1 +1)+n(I 1 +1)(I 2+1) :=  hi (ξ 1 )hj (ξ 2 )hk (ξ 3 )hl (ξ 1 )hm (ξ 2 )hn (ξ 3 ) dξ 1 dξ 2 dξ 3 . ref

The associated algebraic dual polynomial representations, or simply dual polynomials, are linear combinations of the polynomial basis functions, or simply primal polynomials, defined in the previous section, !

" 1 2 3 1 2 3 h 0,0,0 (ξ , ξ , ξ ), · · · , h I 1 ,I 2 ,I 3 (ξ , ξ , ξ ) " ! := h0 (ξ 1 )h0 (ξ 2 )h0 (ξ 3 ), · · · , hI 1 (ξ 1 )hI 2 (ξ 2 )hI 3 (ξ 3 ) M−1 P .

These dual polynomials are always well-defined. This is because the primal polynomials are linearly independent. So the mass matrix MP is injective  and surjective,  1, ξ 2, ξ 3) therefore invertible. Let the finite dimensional space spanned by h (ξ i,j,k I 1 ,I 2 ,I 3

be denoted by 3 P I 1 ,I 2 ,I 3

I 1 ,I 2 ,I 3

. We say 3 P I 1 ,I 2 ,I 3

is the algebraic dual space of the primal

1 2 3 3I ,I ,I

space P . Note that P and P actually represent the same space. The change of basis functions only leads to a different representation. Therefore, we 3 P be the mass matrix also call the algebraic dual space a dual representation. Let M 1 ,I 2 ,I 3 I of 3 P , we can easily see that 3 P MP = I, M

(5)

where I is the identity matrix. Similarly, we can derive the algebraic dual space 1 2 3 3 L and ML be their mass matrices, we 3I ,I ,I of the primal space LI 1 ,I 2 ,I 3 . Let M L have 3 L ML = I. M If ρ h ∈ LI

1 ,I 2 ,I 3

(6)

, σ h , whose vector of expansion coefficients σ satisfies σ = ML ρ,

I will be the representation of ρ h in the algebraic dual space 3 L

(7) 1 ,I 2 ,I 3

.

208

Y. Zhang et al. 1

2

3

To explain how the algebraic dual space of the trace space PI ,I ,I is derived, 2 3 2 3 we take PI−1,I as example. We already know that PI−1,I is a space spanned by   primal polynomials h0 (−1)hj (ξ 2 )hk (ξ 3 ) . With these primal polynomials, we can compute its mass matrix, denoted by Mb . The dual polynomials are then computed by !

" 2 3 2 3 h 0,0,0 (−1, ξ , ξ ), · · · , h 0,I 2 ,I 3 (−1, ξ , ξ ) " ! = h0 (−1)h1 (ξ 2 )h1 (ξ 3 ), · · · , h0 (−1)hI 2 (ξ 2 )hI 3 (ξ 3 ) M−1 b .

  2 3 2 3 The algebraic dual space 3 PI−1,I is spanned by dual polynomials h 0,j,k (−1, ξ , ξ ) . The algebraic dual space of the trace space PI 2

3

1 2 3 3 PI ,I ,I = 3 PI−1,I ∪ 3 PI1

2 ,I 3

1

3

1 ,I 2 ,I 3

∪3 PI−1,I ∪ 3 PI1

eventually can be written as

1 ,I 3

1

2

∪3 PI−1,I ∪ 3 PI1

1 ,I 2

.

I 1 ,I 2 ,I 3

L can be done with the help of the boundary value The divergence of σ h ∈ 3 1 ,I 2 ,I 3 h I 3 . With vector proxies, it can be written as σˆ ∈ P div σ h = NT σˆ h − ET σ h .

(8)

A detailed introduction of algebraic dual polynomial spaces is given in [9].

2.3 Function Spaces in Curvilinear Domains So far, all polynomial spaces are defined only in the reference domain ref |ξ 1 ,ξ 2 ,ξ 3 = [−1, 1]3. Consider an arbitrary domain  and a C1 diffeomorphism  : ref |ξ 1 ,ξ 2 ,ξ 3 → |x 1 ,x 2 ,x 3 . In , the primal polynomials change. Therefore, the mass matrices will also change. But the process of constructing dual polynomials does not change. And as we mentioned before, the metric-independent incidence matrix E and the matrix N remain the same. The way of converting polynomials in Cartesian domain into those in curvilinear domains follows the general coordinate transformation process, for example, see [16]. From now on, notations mentioned in this section not only refer to the reference domain ref , but also refer to the physical domain .

Discrete Equivalence of Adjoint Neumann–Dirichlet Equations

209

3 Weak Formulations 3.1 Discrete Neumann Problem With integration by parts, we can derive the weak formulation of the Neumann problem, (1), written as: For given σˆ ∈ H −1/2(∂), find ω ∈ H 1 () such that     ¯ L2 = tr ω, ¯ σˆ , ∀ω¯ ∈ H 1 (). (9) grad ω, grad ω¯ L2 + (ω, ω) Note that on the right hand side, we use ·, · to represent the duality pairing between 1 2 3 tr ω¯ ∈ H 1/2(∂) and σˆ ∈ H −1/2 (∂). We use finite dimensional space PI ,I ,I 1 2 3 to approximate the space H 1 () and use the algebraic dual trace space 3 PI ,I ,I to approximate the space H −1/2(∂). Then we obtain   grad ωh , grad ω¯ h   ωh , ω¯ h

L2

= ω¯ h,T ET ML E ωh ,

= ω¯ h,T MP ωh ,

L2

and  tr ω¯ h σˆ h d = ω¯ h,T NT σˆ h , ∂

which eventually leads to the discrete formulation of (9), ET ML E ωh + MP ωh = NT σˆ h .

(10)

3.2 Discrete Dirichlet Problem For the Dirichlet problem, (2), the weak formulation is given as: For given σˆ ∈ H −1/2(∂), find σ ∈ H (div, ), tr σ = σˆ such that (div σ , div σ¯ )L2 + (σ , σ¯ )L2 = 0, 1

2

∀σ¯ ∈ H0 (div, ).

3

(11)

I ,I ,I We use algebraic dual space 3 L to approximate H (div, ). With σˆ h ∈ 1 2 3 3 PI ,I ,I given and (8), we obtain     3 P NT σˆ h − ET σ h , div σ h , div σ¯ h 2 = −σ¯ h,T EM L

and

  σ h , σ¯ h

L2

3 L σ h. = σ¯ h,T M

210

Y. Zhang et al.

Therefore, the discrete formulation of (11) is written as 3 P ET σ h + M 3 L σ h = EM 3 P NT σˆ h . EM

(12)

3.3 Equivalence Between Discrete Formulations Now it is time to check if the equivalence between (1) and (2) holds at the discrete level. In other words, it is time to check if the statement that ωh solves (10) if and only if σ h = grad ωh solves (12) is correct. From (4) and (7), we know that σ h , σ h = ML E ωh ,

(13)

is the vector representation of grad ωh in the dual space. If we insert (13) into (12), we obtain 3 L ML E ωh = EM 3 P NT σˆ h . 3 P ET ML E ωh + M EM

(14)

From (10), we know that ET ML E ωh = −MP ωh + NT σˆ h .

(15)

By inserting (15) into (14), we get   3 L ML E ωh = EM 3 P NT σˆ h . 3 P −MP ωh + NT σˆ h + M EM

(16)

From (5) and (6), we know that (16) holds, the equivalence. which proves h h = σ should also be If the equivalence holds, relation ω 1 H ()

H (div,)

satisfied. To prove this, we have 2 h σ

H (div,)

T    3 L σ h + NT σˆ h − ET σ h M 3 P NT σˆ h − ET σ h = σ h,T M

(8)

(13)

=



ML E ωh

T

  3 L ML E ωh M

   + NT σˆ h − ET ML E ωh

T

   3 P NT σˆ h − ET ML E ωh M

3 P MP ωh = ωh,T ET ML E ωh + ωh,T MP M 2 = ωh 1 , (10)

H ()

where we constantly use (5) and (6) and the fact that mass matrices are symmetric.

Discrete Equivalence of Adjoint Neumann–Dirichlet Equations

211

4 Numerical Test Consider the mapping  which maps the Cartesian reference domain ref |ξ 1 ,ξ 2 ,ξ 3 := [−1, 1]3 into the physical domain |x 1 ,x 2 ,x 3 = [0, 1]3 by ⎞ ⎛ @ 1 1 sin(πξ j )⎠ , i = 1, 2, 3. x i = + ⎝ξ i + c 2 2 j

When the deformation coefficient c = 0, the domain  is Cartesian. Otherwise the domain is curvilinear, meaning that a curvilinear coordinate system parametrizes . Examples of such curvilinear domains in R2 are shown in Fig. 1. A manufactured solution of the Neumann problem, (1), is 1

2

3

ωexact = ex + ex + ex .   1 2 3 T solves the Dirichlet problem, (2). Clearly, σ exact = grad ωexact = ex , ex , ex In the domains of different deformation coefficient c, with the boundary condition σˆ = tr σ exact imposed, we solve the discrete formulations (10) and (12) using 1 2 3 Gauss–Lobatto–Legendre (GLL) polynomial spaces  of degree I = I = I = N. The results of the L2 -error of σ h − grad ωh are shown in Fig. 2 (Left) where we can see that the relation σ h = grad ωh is preserved up to the machine precision. With the growth of the polynomial degree, the error increases slowly because of the accumulation of the machine error as the amount of degrees of freedom grows significantly. h In Table 1, the results of the H 1 -norm (div)-norm of σ h are of ω and H h h presented. It is shown that the relation ω 1 = σ holds for all H ()

H (div,)

polynomial degrees irrespective of whether we use the Cartesian domain, c = 0, or

Fig. 1 Curvilinear domains for c = 0.15 (Left) and c = 0.3 (Right) in R2 . The gray lines illustrate the coordinate lines

Y. Zhang et al. H(div)−error

212

10−12

c=0 c = 0.15 c = 0.3

10−13

H 1−error

10−11

= σh

10−10

101

c=0 c = 0.15 c = 0.3

10−2 10−5 10−8 10−11

ωh

σ h − grad ω h

L2−norm

10−9

10−14 2 4 6 8 10 12 14 16 18 20 8 10 12 14 16 18 20 N N   Fig. 2 The L2 -error of σ h − grad ωh (Left) and the p-convergence of the H 1 -error of ωh (Right) for N = 2, 4, · · · , 20 and c = 0, 0.15, 0.3 10−14

2

4

6

Table 1 The H 1 -norm of ωh and H (div)-norm of σ h for polynomial degree N = 2, 4, · · · , 20 and deformation coefficient c = 0, 0.15, 0.3

N 2 4 6 8 10 12 14 16 18 20

c=0 h ω

h σ

6.0720702909 6.0730653395 6.0730653668 6.0730653668 6.0730653668 6.0730653668 6.0730653668 6.0730653668 6.0730653668 6.0730653668

6.0720702909 6.0730653395 6.0730653668 6.0730653668 6.0730653668 6.0730653668 6.0730653668 6.0730653668 6.0730653668 6.0730653668

H1

H (div)

c = 0.15 h ω 1

h σ

5.8899445673 6.0567452129 6.0729332275 6.0730647051 6.0730653557 6.0730653665 6.0730653667 6.0730653668 6.0730653668 6.0730653668

5.8899445673 6.0567452129 6.0729332275 6.0730647051 6.0730653557 6.0730653665 6.0730653667 6.0730653668 6.0730653668 6.0730653668

H

H (div)

c = 0.3 h ω 1

h σ

6.7381947027 5.8849807780 6.0721137212 6.0730525346 6.0730648440 6.0730653428 6.0730653663 6.0730653667 6.0730653668 6.0730653668

6.7381947027 5.8849807780 6.0721137212 6.0730525346 6.0730648440 6.0730653428 6.0730653663 6.0730653667 6.0730653668 6.0730653668

H

H (div)

curvilinear domains, c = 0.15, 0.3. It is also seen that the results always converge to = 6.0730653668. The p-convergence the analytical value ωexact H 1 = σ h H (div)

for the H 1 -error of ωh , therefore also for the H (div)-error of σ h , is shown in Fig. 2 (Right), which shows the exponential convergence of the method.

5 Conclusions By constructing and using primal polynomial spaces and their algebraic dual representations both in the domain and on the boundary, we successfully preserve the equivalence of the div-grad Neumann problem and the grad-div Dirichlet problem at the discrete level in 3-dimensional curvilinear domains. This suggests the further usage of these spaces to structure-preserving methods and hybrid methods.

Discrete Equivalence of Adjoint Neumann–Dirichlet Equations

213

References 1. Bochev, P.B., Hyman, J.M.: Principles of mimetic discretizations of differential operators. In: Compatible Spatial Discretizations, pp. 89–119. Springer, New York (2006) 2. Brezzi, F., Fortin, M.: Mixed and Hybrid Finite Element Methods, vol. 15. Springer Science & Business Media, New York (2012) 3. Carstensen, C., Demkowicz, L., Gopalakrishnan, J.: Breaking spaces and forms for the DPG method and applications including Maxwell equations. Comput. Math. Appl. 72(3), 494–522 (2016) 4. Castillo, J.E., Miranda, G.F.: Mimetic Discretization Methods. Chapman and Hall/CRC, London (2013) 5. Evans, J.A., Scott, M.A., Shepherd, K.M., Thomas, D.C., Vázquez Hernández, R.: Hierarchical B-spline complexes of discrete differential forms. IMA J. Numer. Anal. 40(1), 422–473 (2020) 6. Gerritsma, M.: Edge functions for spectral element methods. In: Spectral and High Order Methods for Partial Differential Equations, pp. 199–207. Springer, Berlin (2011) 7. Gerritsma, M., Jain, V., Zhang, Y., Palha, A.: Algebraic dual polynomials for the equivalence of curl-curl problems (2018). arXiv:1805.00114 8. Gerritsma, M., Palha A., Jain, V., Zhang, Y.: Mimetic spectral element method for anisotropic diffusion. In: Numerical Methods for PDEs. Springer SEMA SIMAI Series, vol. 15, pp. 31–74. Springer, Berlin (2018) 9. Jain, V., Zhang, Y., Palha, A., Gerritsma, M.: Construction and application of algebraic dual polynomial representations for finite element methods (2017). arXiv:1712.09472 10. Jain, V., Zhang, Y., Fisser J., Palha, A., Gerritsma, M.: A conservative hybrid method for Darcy flow (ICOSAHOM 2018, accepted) 11. Kreeft, J., Gerritsma, M.: Mixed mimetic spectral element method for Stokes flow: a pointwise divergence-free solution. J. Comput. Phys. 240, 284–309 (2013) 12. Palha, A., Rebelo, P.P., Hiemstra, R., Kreeft, J., Gerritsma, M.: Physics-compatible discretization techniques on single and dual grids, with application to the Poisson equation of volume forms. J. Comput. Phys. 257, 1394–1422 (2014) 13. Pian, T.H.: Derivation of element stiffness matrices by assumed stress distributions. AIAA J. 2(7), 1333–1336 (1964) 14. Pian, T.H., Tong, P.: Basis of finite element methods for solid continua. Int. J. Numer. Methods Eng. 1(1), 3–28 (1969) 15. Rapetti, F.: High order edge elements on simplicial meshes. ESAIM: Math. Model. Numer. Anal. 41(6), 1001–1020 (2007) 16. Steinberg, S.: Fundamentals of Grid Generation. CRC Press, Boca Raton (1993)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

A Conservative Hybrid Method for Darcy Flow Varun Jain, Joël Fisser, Artur Palha, and Marc Gerritsma

1 Introduction Hybrid formulations [1, 3, 10] are classical domain decomposition methods which reduce the problem of solving one global system to many small local systems. The local systems can then be efficiently solved independently of each other in parallel. In this work we present a hybrid mimetic spectral element formulation to solve Darcy flow. We follow [8] which render the constraints on divergence of mass flux, the pressure gradient and the inter-element continuity metric free. The resulting system is extremely sparse and shows a reduced growth in condition number as compared to a non-hybrid system. This document is structured as follows: In Sect. 2 we define the weak formulation for Darcy flow. The basis functions are introduced in Sect. 3. The evaluation of weighted inner product and duality pairings are discussed in Sect. 4. In Sect. 5 we discuss the formulation of discrete algebraic system. In Sect. 6 we present results for a test case taken from [7].

V. Jain () · J. Fisser · A. Palha · M. Gerritsma Faculty of Aerospace Engineering, TU Delft, Delft, The Netherlands e-mail: [email protected]; [email protected]; [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_16

215

216

V. Jain et al.

2 Darcy Flow Formulation For  ∈ Rd , where d is the dimension of the domain, the governing equations for Darcy flow, are given by, ⎧ ⎨ u + A ∇p = 0 ⎩∇ · u

=f

⎧ ⎪ ⎪ ⎪ ∂ = D ∪ N ⎨ p = pˆ on D , ⎪ ⎪ ⎪ ⎩ u · n = uˆ n on N

in  and

where, u is the velocity, p is the pressure, f the prescribed RHS term, A is a d × d symmetric positive definite matrix, pˆ and uˆ n are the prescribed pressure and flux boundary conditions, respectively.

2.1 Notations   For f, g ∈ L2 (), f, g  denotes the usual L2 -inner product. For vector-valued functions in L2 we define the weighted inner product by, (u, v)A−1 , =

 

 u, A−1 v d ,

(1)



where (· , ·) denotes the pointwise inner product. Duality pairing, denoted by ·, · , is the outcome of a linear functional on L2 () acting on elements from L2 (). Let K be a disjoint partitioning of  with total number of elements K, and Ki is any elementin K , such that, Ki ∈ K . We define the following  A A broken Sobolev spaces [2], H div; K = i H div; Ki , and H 1/2 (∂K ) = i H 1/2 (∂Ki ).

2.2 Weak Formulation The Lagrange functional for Darcy flow is defined as,   L u, p, λ; f =

;

  p ∇ · u − f dK . ; ; ;   + ∂K \D λ (u · n) d + D pˆ (u · n) d − N λ uˆ n d 1 2 K

uT A−1 u dK −

;

K

A Conservative Hybrid Method for Darcy Flow

217

The variational problem is then given by: For given f ∈ L2 (K ), pˆ ∈ H 1/2(ΓD ) 1 and uˆ n ∈ H −1/2(ΓN ) find u ∈ H (div; K ), p ∈ L2 (K ), λ ∈ H 2 (∂K ), such that, ⎧       ⎪ ⎪ (v, u)A−1 ,K − ∇ · v, p  + (v · n) , λ ∂ \ = − v · n, pˆ  ⎪ K K D D ⎪ ⎨     − q, ∇ · u  = − q, f  K K ⎪ ⎪ ⎪     ⎪ ⎩ μ, (u · n) = μ, uˆ n  ∂ \ K

D

N

∀ v ∈ H (div; K ) ∀ q ∈ L2 (K )

.

1 2

∀ μ ∈ H (∂K )

(2)

3 Basis Functions 3.1 Primal and Dual Nodal Degrees of Freedom Let ξj , j = 0, 1, . . . , N, be the N + 1 Gauss–Lobatto–Legendre (GLL) points in I ∈ −1, 1 . The Lagrange polynomials hi (ξ ) through ξj , of degree N, given by, 

  hi ξ =

   ξ 2 − 1 LN ξ    , N (N + 1) LN ξi ξ − ξi

form the 1D primal nodal polynomials which satisfy, hi (ξj ) = δij . Let a h and b h be two polynomials expanded in terms of hi ξ . The L2 —inner product is then given by,   a h , b h = aT M(0) b , I

(0)

where Mi,j =



1 −1

hi (ξ ) hj (ξ ) dξ ,

  and, a = [a0 a1 . . . aN ] and b = b0 b1 . . . bN are the nodal degrees of freedom. We define the algebraic dual degrees of freedom,  a, such that the duality pairing is simply the vector dot product between primal and dual degrees of freedom, 

a h , bh

 I

= aT b := aT M(0) b

⇒

 a = M(0) a .

Thus, the dual degrees of freedom are linear functionals of primal degrees of freedom.

218

V. Jain et al.

3.2 Primal and Dual Edge Degrees of Freedom   The edge polynomials, for the N edges between N + 1 GLL points ξj −1 , ξj , of polynomial degree N − 1, are defined as [4], ej (ξ ) = −

j −1  dhk k=0

dξ

 (ξ ) ,

ξj

ei (ξ ) = δij .

such that ξj−1

Let ph and q h be two polynomials expanded in edge basis functions. The inner product in L2 space is given by,   ph , q h = pT M(1) q , I

where M(1) i,j =



1 −1

ei (ξ ) ej (ξ ) dξ ,

    and, p = p1 p2 . . . pN and q = q1 q2 . . . qN are the edge degrees of freedom. As before, we define the dual degrees of freedom such that,   ph , q h =  pT q := pT M(1) q I

⇒

p = M(1) p . 

A similar construction can be used for dual degrees of freedom in higher dimensions. For construction of the dual degrees of freedom in 2D see [8] and for 3D see [9].

3.3 Differentiation of Nodal Polynomial Representation   Let a h ξ be expanded in Lagrange polynomials, then N N        d h  d  a ξ = ai − ai−1 ei ξ . ai hi ξ = dξ dξ i=0

(3)

i=1

Therefore, taking the derivative of a polynomial involves two steps: First, take the difference of degrees of freedom; second, change of basis from nodal to edge [4].

4 Discrete Inner Product and Duality Pairing For 2D domains, the higher dimensional primal basis are constructed using the tensor product of the 1D basis.

A Conservative Hybrid Method for Darcy Flow

219

For the weak formulation (2) we expand the velocity uh in primal edge basis as, N N N  N      uh ξ, η = uxi,j hi (ξ ) ej (η) ıˆ + uy i,j ei (ξ ) hj (η) jˆ , i=0 j =1

(4)

i=1 j =0

; where ux i,j denotes the flux, u · n, over the vertical edges and uy i,j the flux over the horizontal edges, see Fig. 1.

4.1 Weighted Inner Product Using (1) and the expansions in (4), the weighted inner product is evaluated as,   v h , uh

A−1 ,K

=



(1) vTKi MA −1 ,K uKi , i

Ki

Fig. 1 Discretized domain for K = 3 × 3, N = 3. The blue dots represent the pressure boundary condition p, ˆ and the blue edges represent the velocity boundary condition uˆ n

220

V. Jain et al.

where, uKi are the degrees of freedom in element Ki , and = M(1) A−1 ,K i

⎞T

⎛



⎝

hi (ξ ) ej (η) ei (ξ ) hj (η)

Ki

⎠ A

 −1



⎞

⎛

ξ, η ⎝

hi (ξ ) ej (η) ei (ξ ) hj (η)

⎠ dKi .

For mapping of elements please refer to [6].

4.2 Divergence of Velocity Divergence of velocity, ∇ · uh , is evaluated using (3), but now for 2D, ∇ · uh = =

∂ ∂x

1 (cf. [10, Chapter 8]). We say that * ⊂ R2 is a regular Jordan arc of class Cm , for m ∈ N, if it is the image of a bijective parametrization, denoted by r = (r1 , r2 ), such that its components are Cm (B  → * and r (t)2 > 0, ∀ t ∈ B  , where ·2 is the  )-functions, r : B Euclidean norm. Similarly, we define ρ-analytic arcs as those whose components are ρ-analytic. Throughout, we will assume that for any * regular Jordan arc, there ˜ which is a closed and keep the same regularity. exists an extension of * to *, Consider a finite number M ∈ N of at least C1 -arcs, written {i }M i=1 , such that their closures are mutually disjoint. Moreover, we assume that there are disjoint domains i whose boundaries are given by extensions ∂i = 3 i , for i = 1, . . . , M. Let us define  :=

M E

i

and

 := R2 \ .

i=1

We say that  is of class Cm , m ∈ N, if each arc i is of class Cm and analogously for the ρ-analytic case. For i ∈ {1, . . . , M}, let ri : B  → i and gi :  i → C. We claim that g = (g1 , . . . , gM ) is of class Cm () if gi ◦ ri ∈ Cm (B ), for i ∈ {1, . . . , M}. A similar definition holds for the analytic case. Let G ⊆ Rd , d = 1, 2, be an open domain. For s ∈ R, we denote by H s (G) s the standard Sobolev spaces, by Hloc (G) their locally integrable counterparts [8, −s 3 Section 2.3], and by H (G) the corresponding dual spaces. The corresponding duality product (when the dual space of L2 (G) is identified with itself) is denoted 3s (G) refers to mean-zero spaces [5, Section 2.3]. We will also ·, ·G . Finally, H 0 make use of the following Hilbert space in R2 :

W (G) :=

⎧ ⎪ ⎨ ⎪ ⎩

U ∈ D∗ (G) : *

U (x) 1 + x22 log(2 + x22 )

⎫ ⎪ ⎬

∈ L2 (G), ∇U ∈ L2 (G) , ⎪ ⎭

where D∗ (G) is the dual space of C∞ (G) = ∩n>1 Cn (G). For s ∈ R and for the finite union of disjoint open arcs , we define Cartesian product spaces as Hs () := H s (1 ) × H s (2 ) × · · · × H s (M ).

Spectral Galerkin Method for Dirichlet Problems on Multiple Open Arcs

385

3s () and H 3s () are defined similarly. Also, Hs (B Spaces H ) is to be understood as 0 AM s the Cartesian product i=1 H (B  ). Finally, given an open bounded neighborhood Gi such that i ⊂ ∂Gi , Dirichlet traces are defined as extensions to H s (Gi ), for s ≥ 1/2, of the following operator (applied to smooth functions): γi± u(y) := lim u(y ± #ni (y)), #↓0

 (t), −r  (t)) and t such that where ni (y) is the unitary vector with direction (ri,2 i,1 r(t) = y. For a function u defined in an open neighborhood of i such that γi+ u = γi− u, we denote γi u := γi± u. 1

1 () Problem 1 (Volume Problem) Let g ∈ H 2 () and κ ≥ 0. We seek U ∈ Hloc such that

− U − κ 2 U = 0

in ,

(1)

γi± U = gi

for i = 1, . . . , M,

(2)

Condition at infinity(κ).

(3)

The behavior at infinity (3) depends on κ in the following way: if κ > 0, we employ the classical Sommerfeld condition [8, Section 3.9]. If κ = 0, we seek for solutions U ∈ W (). This last condition was discussed in detail in [5, Remarks 3.9, 4.2 and 4.5] with uniqueness proofs for κ ≥ 0 provided in [5, Propositions 3.8 and 3.10]. For κ ≥ 0, we can express U solution of Problem 1 as U (x) =

M 

(SLi [κ]λi )(x),

∀ x ∈ ,

(4)

i=1

where

 (SLi [κ]λi )(x) :=

Gκ (x, y)λi (y)di (y),

∀ x ∈ ,

i

denotes the single layer potential generated at a curve i with Gκ the corresponding fundamental solution, defined as in [8, Section 3.1]. It is direct from (4) that U solves (1)–(2) in  (see [8, Theorem 3.1.1]). Also, it displays the desired behavior at infinity as long as each λi lies in the right functional space [5, Section 4]. In order to find the surface densities λi , we take Dirichlet traces γi± of the SLj and impose boundary conditions (2). This naturally defines of weakly singular boundary integral operators: Lij [κ] :=

 1 + γi SLj [κ] + γi− SLj [κ] = γi SLj [κ], 2

and an equivalent boundary integral equation problem to Problem 1.

386

C. Jerez-Hanckes and J. Pinto 1

Problem 2 (Boundary Integral Problem) Let g ∈ H 2 (). For κ > 0, we seek 3− 12 () such that λ = (λ1 , . . . , λM ) ∈ H L[κ]λ = g, 3− 12 () → H 12 () is a matrix operator with entries L[κ]ij = Lij [κ], where L[κ] : H 1

3− 2 (), given g in the dual space of for i, j ∈ {1, . . . M}. If κ = 0, we seek λ ∈ H 0 the aforementioned space. Theorem 1 (Theorem 4.13 in [5]) For κ > 0, Problem 2 has a unique solution 1 3− 12 (), whereas for κ = 0 a unique solution exists in the subspace H 3− 2 (). λ∈H 0 Also, the following continuity estimate holds λ3− 1 H

2 ()

≤ C(, κ)g

1

H 2 ()

.

3 Spectral Discretization 3− 21 () that can be used We present a family of finite dimensional subspaces in H to approximate the solution of Problem 2 (cf. [4, 6]). Let TN (B ) denote the space spanned by first kind Chebyshev polynomials, denoted by {Tn }N n=0 , of degree lower B or equal than N on √  , orthogonal with the L2 (−1, 1) inner product, under the weight w−1 with w(t) := 1 − t 2 . Now, let us construct elements pni = Tn ◦ r−1 i over each arc i spanning the space TN (i ). For practical reasons, we define the normalized space:

TN (i ) :=

⎧ ⎪ ⎨ ⎪ ⎩

p¯ i ∈ C(i ) : p¯ni :=

pni  ri ◦ r−1 i 2

⎫ ⎪ ⎬

,

pni ∈ TN (i ) . ⎪ ⎭

We account for edge singularities by multiplying the basis {p¯ ni }N n=0 by a suitable weight:   QN (i ) := qni := wi−1 p¯ ni : p¯ni ∈ TN (i ) , wherein wi := w ◦ r−1 i . The corresponding basis for QN (i ) will be denoted . By Chebyshev orthogonality, we can easily define the mean-zero subspace {qni }N n=0 QN,0 (i ) := QN (i ) \ Q0 (i ), spanned by {qni }N n=1 . With these definitions, we

Spectral Galerkin Method for Dirichlet Problems on Multiple Open Arcs

387

set the discretization space for a Galerkin-Bubnov solution of Problem 2 as ⎧ ⎨AM Q N,0 (i ) HN [κ] := Ai=1 ⎩ M QN (i ) i=1

for κ = 0, for κ > 0. 1

Problem 3 (Linear System) For κ > 0, let N ∈ N and g ∈ H 2 () be the same as in Problem 2. Then, we seek coefficients u = (u1 , . . . , uM ) ∈ CM(N+1) , such that L[κ]u = g. Therein, we have defined the Galerkin matrix L[κ] ∈ CM(N+1)×M(N+1) composed of matrix blocks Lij [κ] ∈ C(N+1)×(N+1) whose entries are   j (Lij [κ])lm = Lij [κ]qm , qli

i

  Bij [κ]w−1 Tm , w−1 Tl . = L B 

Bij [κ] is the weakly-singular operator whose kernel is parametrized by ri , rj There, L and right-hand g = (g1 , . . . , gM ) ∈ CM(N+1) with components   (gi )l = gi , qli

i

  = B gi , w−1 Tl , B 

where B gi = gi ◦ ri . The approximation λN ∈ HN [κ] is constructed as (λN )i =

N 

i (ui )m qm in i ,

for all i ∈ {1, . . . , M}.

m=0

For k = 0 we need g as in Problem 2; we also have u ∈ CMN , and L[0] ∈ CMN×MN since the approximation space is HN [0]. By conformity and density of these spaces 3− 12 (), one derives the following result: in H Theorem 2 (Theorem 4.23 [4]) Let κ ≥ 0, m ∈ N with m > 2,  ∈ Cm , g ∈ Cm (), and λ be the only solution of Problem 2. Then, there exists N0 ∈ N such that for every N > N0 ∈ N there is a unique λN ∈ HN [κ] solution of Problem 3. Moreover, the following error convergence rates hold λ − λN 

1

3− 2 () H

≤ C(, κ)N −m+1 .

Moreover, if  and g are ρ-analytic with ρ > 1, we have the following superalgebraic convergence rates λ − λN 

1 3− 2

H

()

√ ≤ C(, κ)ρ −N+2 N,

where C(, κ) is a positive constant, which does not depend on N.

388

C. Jerez-Hanckes and J. Pinto

Remark 1 Observe that the constants C(, κ) and N0 depend on the geometry and frequency. To the best of our knowledge previous convergence results for 2D arcs are somehow limited. For intervals, the result was established in [6] whereas for more general arc results are only obtained for the Laplace case [1]. Super-algebraic convergence rates can be achieved by the method detailed in [3], though their scheme is limited to intervals and to the case of elliptic problems (N0 = 0). More complex cases are still an open problem.

4 Numerical Implementation and Compression Algorithm Before fleshing out our proposed compression technique, we explain how L[κ] and g of Problem 3 are computed. For the right-hand side, one must compute integrals of the form: 

1 −1

B g (t)w−1 (t)Tl (t)dt,

∀ l ∈ N0 ,

which corresponds to Fourier-Chebyshev coefficients of B g (t) and can be approximated using the Fast Fourier Transform [10]. Computations for matrix terms Lij [κ] are split into two groups: (a) cross-interactions, where test and trial functions supports lie along curves i , j with i = j ; and (b) self-interactions, where both trial and test functions are defined on the same curve. As for cross-interactions the integral kernel is smooth, we use the same computational procedure for the righthand side. For self-interactions, the kernel function has a singularity that can be characterized as Gk (r(t), r(s)) = (2π)−1 log |t − s|J0 (kr(t) − r(s)2 ) + Gr (t, s),

t = s,

for t, s ∈ B , where J0 is the zeroth-order first kind Bessel function, and Gr is a regular function. Thus, integration for the regular part is done as in the crossinteraction case, while integrals with the first term as kernel are obtained by convolution as integrals for log |t − s| are known (see [6, Remark 4.2]). Yet, as κ increases, larger values of N will be required, and thus, the need to compress the resulting matrix terms. As stated in [10, Chapters 7 and 8], the regularity of a function controls the decay of its Fourier-Chebyshev coefficients. Hence, as the entries of the matrix L[κ] are precisely such coefficients, for a smooth kernel one observes fast decaying terms. This implies that we can select small blocks to approximate the matrix and obtain a sparse approximation by discarding the remaining entries, based on a predetermined tolerance # > 0. Specifically, the kernel function is smooth when we compute cross-interactions. Let the routine Quadrature(l,m) compute the term (l, m) of this interaction matrix using a 2D

Spectral Galerkin Method for Dirichlet Problems on Multiple Open Arcs

389

Gauss-Chebyshev quadrature. Given a tolerance # > 0, we minimize the number of computations needed by performing the following binary search: Matrix Compression Algorithm

INPUT: Tolerance (Tol), Max level of search (Lmax) OUTPUT: Number of columns to use (Ncols) INITIALIZE: Ncols = N, level = 0, a = 0, b = N While{level < Lmax} m = (a+b)/2 Tleft = m-1 Tcenter = m Tright = m+1 Veft = abs(Quadrature(0,Teft)) Vcenter = abs(Quadrature(0,Tcenter)) Vright = abs(Quadrature(0,Tright)) If{Vright & Vcenter < 0.5*Tol} or {Vleft & Vcenter < 0.5*Tol} b = m Else a = m EndIF level++ EndWhile Ncols = b

The algorithm returns the minimum number of columns required, Ncols , by searching in the first row the minimum index such that the matrix entries’ absolute value is lower than #. The binary search is restricted to a depth Lmax ∈ N. The same procedure is used to estimate the number of rows, Nrows , by executing a binary search in the first column. Once Ncols and Nrows are selected, we define N# := max{Nrows , Ncols } and compute the block of size N# × N# as in the full matrix implementation. The matrix compression percentage will strongly depend on the regularity of the arcs involved. For ρ-analytic arcs, using [10, Theorem 8.1] we can prove the lower bound: N# ≥

− log # , 2ϒ log ρ

where ϒ is an upper bound for the absolute value of the kernel in the corresponding Bernstein ellipse. However, since compression is done by a binary search, the bound for the compression rate depends on Lmax as N# ≥

N . 2Lmax

390

C. Jerez-Hanckes and J. Pinto

Compression of self-interaction blocks does not follow the same ideas. In fact, these blocks can be characterized as two perturbations over the canonical case,  = B  for κ = 0, leading to a diagonal matrix. Namely, these are  2→ , similar to the 1. A low frequency perturbation caused by the mapping ri : B cross-interaction case. 2. A frequency perturbation that creates banded matrices. In order to reduce memory consumption—though not computational time—we discard the entries of the self-interaction matrices lower than the given tolerance. As expected, matrix compression induces an extra error as it perturbs the original linear system solved by λN in Problem 3. We denote by L# [k] the matrix generated by the compression algorithm with tolerance #, and define the matrix difference

L# [k] := L# [k] − L[k]. We seek to control the solution u# = u + u of (L[k] + L# [k])u# = g, where u and g are the same as in Problem 3. In order to bound this error, we will assume that, for every pair of indices (i, j ) in the matrix L[k], we have, |( L# [k])ij | < #.

(5)

Theorem 3 Let N ∈ N be such there is only one λN solution of Problem 3. Then, there is a constant C(, κ) > 0, not depending on N, such that     u2  N# . ≤ u2 C(κ, ) − N#  Proof By [7, Section 1.13.2] we have that L# [k]  u2 2 , ≤ (L[k])−1 − L# [k] u2 2 2 and thus, we need to estimate L# [k] 2 and (L[k])−1 . The bound for the first 2 term is direct from (5) and matrix norm definitions. By the classical bound of a matrix inverse and the continuity of the associated boundary integral operator, it holds that −1 L[k]−1 g ≥ L[k]g 2 ≥ C(κ, Γ ), 2

from where the result follows directly.

& %

Spectral Galerkin Method for Dirichlet Problems on Multiple Open Arcs

391

We can also estimate the error introduced by the compression algorithm in terms of < # i the energy norm. In order to do so, define (λ#N )i := N (u m=0 i )m qm in i . By the same arguments in the above proof, we obtain λN − λ# 1 N 3− H

2 ()

≤ C1 (κ, ) g

1 H() 2

#N 3/2 , C2 (κ, ) − #N

where g is the same that in Problem 2 and C1 (κ, Γ ), C2 (κ, Γ ) are two different constants. Remark 2 Our compression algorithm produces a faster and less memory demanding implementation of the spectral Galerkin method at the cost of accuracy loss, similar to fast multipole or hierarchical matrices methods. Moreover, once we have compressed the matrix, we can implement a fast matrixvector product.

5 Numerical Results To illustrate the above claims, Fig. 1 presents convergence results for different wavenumbers, κ = 0, 25, 50, 100 for a configuration of M = 28 arcs. As the chosen geometry and excitation are given by analytic functions, Theorem 2 predicts exponential rate of convergence as observed numerically. Table 1 provides matrix compression results for κ = 100 and for the same geometry of Fig. 1. It presents the percentage of non-zero entries (%NNZ) and relative errors as bounded in Theorem 3 as functions of the maximum level of binary

Error

100

10-5

10-10

0

50

100

150

200

250

300

N

(a) Geometry

(b) Convergence

− 21

(Γ)-norm

Fig. 1 (a) Smooth geometry with M = 28 open arcs parametrized as ri (t) = (ai t, ci sin(bi t)+di ), with ai ∈ [0.14, 0.25], bi ∈ [0, 0.2], ci ∈ [1, 2], di ∈ [0, 20], t ∈ [−1, 1]. (b) Convergence results for different wavenumbers and a planewave excitation along (1, 1). Errors computed against an overkill solution using N = 660 per arc

392

C. Jerez-Hanckes and J. Pinto

Table 1 Compression performance for κ = 100 Order 5 10 20 40 60 80 5 10 20 40 60 80

Lmax = 2 % NNZ # = 1e−6 65.24 81.62 89.41 77.63 45.25 27.20 # = 1e−10 65.29 81.68 89.86 83.46 51.94 33.86

Rel. error

Lmax = 3 % NNZ

Rel. error

Lmax = 4 % NNZ

Rel. error

5.05e−01 5.32e−01 2.33e−01 9.10e−04 2.02e−07 1.97e−07

65.24 81.62 88.62 70.63 36.68 21.97

5.05e−01 5.32e−01 2.33e−01 9.10e−04 2.76e−07 3.17e−07

65.24 81.62 88.31 67.11 33.36 19.50

5.05e−01 5.32e−01 2.33e−01 9.10e−04 3.31e−07 3.35e−07

5.05e−01 5.32e−01 2.33e−01 9.10e−04 2.14e−09 2.31e−09

65.29 81.68 89.59 78.70 44.87 26.89

5.05e−01 5.32e−01 2.33e−01 9.10e−04 3.19e−09 1.73e−08

65.29 81.68 89.44 76.28 40.70 23.78

5.05e−01 5.32e−01 2.33e−01 9.10e−04 3.89e−09 1.73e−10

search (Lmax ), tolerances (#), and polynomial order per arc (Order). For low orders (Order < 60), relative errors are quite large, and therefore, most of the matrix terms are kept. This is due to an insufficient number of matrix entries to solve the problem with good accuracy (see Fig. 1), rendering compression pointless. On the other hand, once convergence is achieved, the compression error drastically decreases along with the percentage of matrix terms stored.

References 1. Atkinson, K.E., Sloan, I.H.: The numerical solution of first-kind logarithmic-kernel integral equations on smooth open arcs. Math. Comput. 56(193), 119–139 (1991) 2. Costabel, M., Dauge, M.: Crack singularities for general elliptic systems. Math. Nachr. 235(1), 29–49 (2002) 3. Hewett, D.P., Langdon, S., Chandler-Wilde, S.N.: A frequency-independent boundary element method for scattering by two-dimensional screens and apertures. IMA J. Numer. Anal. 35(4), 1698–1728 (2014) 4. Jerez-Hanckes, C., Pinto, J.: High-order Galerkin method for Helmholtz and Laplace problems on multiple open arcs. Technical Report 2018-49, Seminar for Applied Mathematics, ETH Zürich (2018) 5. Jerez-Hanckes, C., Pinto, J.: Well-posedness of Helmholtz and Laplace problems in unbounded domains with multiple screens. Technical Report 2018-45, Seminar for Applied Mathematics, ETH Zürich (2018) 6. Jerez-Hanckes, C., Nicaise, S., Urzúa-Torres, C.: Fast spectral Galerkin method for logarithmic singular equations on a segment. J. Comput. Math. 36(1), 128–158 (2018)

Spectral Galerkin Method for Dirichlet Problems on Multiple Open Arcs

393

7. Saad, Y.: Iterative Methods for Sparse Linear Systems. Computer Science Series. PWS Publishing Company, Boston (1996) 8. Sauter, S., Schwab, C.: Boundary Element Methods. Springer Series in Computational Mathematics. Springer, Berlin (2010) 9. Stephan, E.P.: A boundary integral equation method for three-dimensional crack problems in elasticity. Math. Methods Appl. Sci. 8(4), 609–623 (1986) 10. Trefethen, L.: Approximation Theory and Approximation Practice. Other Titles in Applied Mathematics. SIAM, Philadelphia (2013) 11. Verrall, G., Slavotinek, J., Barnes, P., Fon, G., Spriggins, A.: Clinical risk factors for hamstring muscle strain injury: a prospective study with correlation of injury by magnetic resonance imaging. Br. J. Sports Med. 35(6), 435–439 (2001)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Explicit Polynomial Trefftz-DG Method for Space-Time Elasto-Acoustics H. Barucq, H. Calandra, J. Diaz, and E. Shishenina

1 Trefftz-DG Formulation for the Elasto-Acoustic Equation Trefftz methods are particular finite element methods where the basis and test functions are locally solutions to the partial differential equation that governs the problem to be solved. Compared to the existing literature for solving frequency problems, space-time Trefftz methods are still not widely used. One reason could be that they require using space-time meshes [6, 12]. To our knowledge, few references on Trefftz approximations of time-dependent wave equations are available and they mainly address theoretical properties in the case of Acoustics and Electromagnetism [4, 8, 10, 11]. They provide convergence and stability studies and some numerical results are displayed by using plane wave bases in 1D + time dimension. Numerical in 2D + time dimensions are proposed in [4] for electromagnetism. There are also some studies devoted to the second-order formulation of the acoustic wave equation approximated in Trefftz spaces by the mean of Lagrange multipliers [1, 13]. In [3], we have proposed a Trefftz-DG formulation for elasto-acoustic. The method required the inversion of a huge sparse matrix. The goal of this paper is to show how to derive a semi-explicit scheme, requiring only the inversion of a block-diagonal matrix on each element of the mesh.

H. Barucq · J. Diaz · E. Shishenina () Magique-3D, Inria, E2S UPPA, CNRS, Université de Pau et des Pays de l’Adour, Pau, France e-mail: [email protected]; [email protected]; [email protected] H. Calandra Total SA, CSTJF Total, Pau, France e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_31

395

396

H. Barucq et al.

In this section, following [10] and the framework therein, we propose a formulation of the elasto-acoustic coupling reading as a first-order system. Here and further the sub-scripts F and S corresponds to the acoustic (fluid) and elastodynamic (solid) domains.

1.1 Elasto-Acoustic Equations We introduce a space-time domain Q ≡ (F ∪ S ) × I , where F ⊂ Rd is a bounded Lipschitz domain of dimension d filled with fluid, S ⊂ Rd is a bounded Lipschitz elastodynamic domain of dimension d filled with solid, and I ≡ [0, T ] is the time interval. All medium parameters cF ≡ cF (x) and ρF ≡ ρF (x), standing for the acoustic wave propagation velocity and fluid density respectively, as well as the inverted stiffness tensor C−1 (x) ≡ A(x) and the solid density ρS ≡ ρS (x), are assumed to be piecewise constant and positive. We denote by F S = F ∩ S the fluid-solid interface. The elasto-acoustic system of equations is based on the coupling of the first-order acoustic equation, written in terms of velocity vF ≡ vF (x, t) and pressure p ≡ p(x, t) fields: ⎧ 1 ∂p ⎪ ⎪ + divvF = f ⎪ ⎪ 2 ⎪ ⎪ cF ρF ∂t ⎪ ⎪ ⎪ ⎨ ∂vF + ∇p = 0 ρF ∂t ⎪ ⎪ ⎪ ⎪ ⎪ vF (·, 0) = vF 0 , p(·, 0) = p0 ⎪ ⎪ ⎪ ⎪ ⎩ vF · nF = gF

in QF , in QF ,

(1)

in F , in ∂F \F S × I,

where nF is the normal vector to ∂F , the source term f ≡ f (x, t), the boundary condition gF , the velocity vF 0 and the pressure p0 are the initial data, with the firstorder elastodynamic system, written in terms of velocity vS ≡ vS (x, t) and stress tensor (symmetrical and positive) σ ≡ σ (x, t) fields: ⎧ ⎪ ∂σ ⎪ ⎪ A − ε(vS ) = 0 ⎪ ⎪ ∂t ⎪ ⎪ ⎪ ⎪ ⎨ ∂vS ρS − div σ = 0 ∂t ⎪ ⎪ ⎪ ⎪ vS (·, 0) = vS 0 , σ (·, 0) = σ 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ σ n = gS S

in QS , in QS , in S , in ∂S \F S × I,

(2)

Explicit Polynomial Trefftz-DG Method for Space-Time Elasto-Acoustics

397

where nS is the normal vector to ∂S , the boundary condition gF , the velocity vS 0 and the stress tensor σ 0 are the initial data. The transmission conditions between the two systems (2) and (1) represent the continuity of velocity and stress normal components F S : ⎧ ⎨ vF · nF S = vS · nF S

at F S ,

⎩ − pn = σ n FS FS

at F S .

(3)

The velocities aligned with the interface and the tangential stress remain unconstrained.

1.2 Space-Time Trefftz-DG Formulation We introduce a non-overlapping space-time mesh Th on Q composed of space-time Lipschitz elements KF ⊂ F × I and KS ⊂ S × I . We denote by TF h (resp. TSh ) the restriction of Th to the fluid (resp. solid) domain. Let nKF ≡ (nxKF , ntKF ) be the outward-pointing unit normal vector on ∂KF , and nKS ≡ (nxKS , ntKS ) be the outward-pointing unit normal vector on ∂KS . We assume that all medium parameters are constant in KF and KS respectively. The mesh skeleton Fh ≡ 8 ∂KF,S can be decomposed into families of the internal FQ h faces, the fluidKF,S ∈Th

0 T solid FFh S faces, the boundary FD h faces, the initial and final time Fh and Fh element faces respectively, as it shown in Fig. =1. We introduce the space Vh (Th>) as a subspace of L2 (Q) defined by Vh (Th ) = φ ∈ L2 (Q), φ|KF,S ∈ Pp (KF,S ) . The unknowns (vF h , ph , vSh , σ h ) are supposed to be in Vh (Th ) ≡ Vh (TF h )d × 2 Vh (TF h ) × Vh (TSh )d × Vh (TSh )d . We consider the test functions ωF , q, ωS , ξ in

T KF

KS

time I

Fig. 1 Example of 1D + time mesh Th covering Q. The internal element faces FQ h are represented by dotted line, the element faces of fluid-solid interface FFh S —by dash-dotted line, the boundary element faces FD h —by thick line, the initial F0h and the final FTh time element faces—by double and dashed line respectively

0

ΩF

ΩS space domain

398

H. Barucq et al.

T(Th ) for vF , p, vS and σ respectively, where the Trefftz space T(Th ) is defined on the mesh Th as follows:  1 ∂q ∂ωF T(Th ) ≡ (ωF , q, ωS , ξ ) ∈ Vh (Th ) s. t. 2 + divωF = 0, ρF + ∇q = 0 ∂t cF ρF ∂t ∀KF ∈ TF h , and A

∂ξ ∂t

− ε(ωS ) = 0, ρS

 ∂ωS − div ξ = 0 ∀KS ∈ TSh . ∂t

This space is of Trefftz type since it is a subspace of the regular space Vh (Th ) composed of local solutions of the volumic governing equations (1) and (2) set in each element KF and KS respectively. As in the standard DG methods, the next step in order to obtain the variational formulation consists in multiplying the equations of (1) by the test functions q and ωF in T(Th ), and the equations of (2) by the test functions ξ and ωS in T(Th ) respectively, and, as is standard in space-time DG methods, we integrate by parts the obtained equations not only in space but also in time:   ! KF ∂K

F

 ! KS ∂K

 KF K

1 cF2 ρF

" p˘h q ntKF + q vˆ F h · nxKF + ρF v˘ F h · ωF ntKF + pˆ h ωF · nxKF ds +

" A σ˘ h : ξ ntKS − ξ vˆ Sh · nxKS + ρS v˘ Sh · ωS ntKS − σˆ h : (ωS ⊗ nxKS ) ds =

S

f qdv.

(4)

F

Thanks to the choice of test functions the left hand side of the above space-time formulation contains only surface integrals. The numerical fluxes in time vˆ F h , pˆh , vˆ Sh , σˆ h and in space v˘ F h , p˘h , v˘ Sh , σ˘ h are defined in the standard DG notations [2, 3, 7] as follows: ⎛ ⎞ ⎛ ⎞ vSh · nxKF + δ1 (σ h nxKF + ph nxKF ) · nxKF vˆ F h · nxKF ⎜ ⎟ ⎜ ⎟ ⎜pˆ h ⎟ ⎜ph + α1 (vF h · nxK − vSh · nxK ) ⎟ F F ⎜ ⎟≡⎜ ⎟ on FF S , h ⎜vˆ ⎟ ⎜v − δ (σ nx + p nx ) ⎟ Sh Sh 1 h h KS ⎝ ⎠ ⎝ ⎠ KS x σˆ n −ph nxKS + α1 (vF h · nxKS − vSh · nxKS )nxKS # h KS $ # $ vˆ F h · nxKF gF ≡ , ph + α1 (vF h · nxKF − gF ) pˆ h # $ # $ vˆ Sh vSh − δ1 (σ h nxKS − gS ) ≡ on FD h. σˆ h nxKS gS

Explicit Polynomial Trefftz-DG Method for Space-Time Elasto-Acoustics

#

# $ # $ # $ {vSh} − δ1 [[σ h ]]x {vF h} + β1 [[ph ]]x vˆ Sh ≡ , ≡ {ξ } − γ1 [[vF h ]]x {ph} + α1 [[vF h ]]x σˆ h # $ # $ # $ # $ v˘ F h {vF h} + α2 [[vF h ]]t v˘ Sh {vSh} + γ2 [[vF h ]]t ≡ , ≡ p˘ h {ph} + β2 [[ph ]]t σ˘ h {σ h} + δ2 [[σ h ]]t # $ # $ # $ # $ v˘ F h v v˘ Sh v ≡ Fh , ≡ Sh p˘ h ph σ˘ h σh $ # $ # ( 1 − α2 )vF h + ( 12 + α2 )vF 0 v˘ F h ≡ 21 , p˘ h ( 2 − β2 )ph + ( 12 + β2 )p0 # $ # $ v˘ Sh ( 12 − γ2 )vSh + ( 12 + γ2 )vS0 ≡ 1 σ˘ h ( 2 − δ2 )σ h + ( 12 + δ2 )σ 0 vˆ F h pˆ h

399

$

on FQ h, Q

on Fh , on FTh ,

on F0h ,

Here, α1 , α2 , β1 , β2 , δ1 , δ2 , γ1 , and γ2 are positive penalty parameters. As in standard DG methods, a suitable choice of these penalty parameters allows one to prove stability of the overall method. It is shown in [2, 3] that they contribute to the accuracy and convergence of the numerical method. We refer to [2, 3] for more details on the definition of the numerical fluxes. Summing the contribution (4) of all elements KF , KS ∈ Th , and introducing  the bilinear AT DG (· ; ·) and the linear T DG ·) forms for the left-hand side and the right-hand side expressions respectively, we obtain the Trefftz-DG formulation for the elasto-acoustic problem: Seek (vF h , ph , vSh , σ h ) ∈ T(Th ) such that, for all (ωF , q, ωS , ξ ) ∈ T(Th ) it holds true:     AT DG (vF h , ph , vSh , σ h ); (ωF , q, ωS , ξ ) = T DG ωF , q, ωS , ξ . (5)

The analysis of well-posedness of (5) is based on the coercivity and continuity estimates of the bilinear and linear forms in mesh-dependent norms [2, 3]. The proof is similar to the one given in [10] where the acoustic wave equation is addressed. In Sect. 2 we provide the algorithm of the Trefftz-DG formulation (5), and we discuss different analytical and numerical approaches for its optimization.

2 Implementation of the Algorithm The numerical implementation of the Trefftz-DG formulation is different from the standard DG ones which address the space and time integration separately. Standard DG space integrations have the interesting feature of leading to a block-diagonal

400

H. Barucq et al.

mass matrix and allow then the use of explicit time integration. The computational costs thus depend on a CFL condition which sets the value of the time step as a function of the space step. On the other hand, a naive implementation of TrefftzDG methods require performing a space-time integration which leads to invert a sparse matrix whose size tends to be huge. It is thus not obvious that a crude implementation of the Trefftz-DG algorithm does not generate additional cost as compared to standard DG ones. In this section we provide some important steps of implementation of Trefftz-DG formulation (5) and discuss optimization techniques. The complete algorithm with more numerical details can be found in [2, 3].

2.1 Change-Over Between the Time Slabs To simplify the presentation, we assume here that we use the same order of f approximation on each cells, so that we have Ndof degrees of freedom on fluid s cells and Ndof degrees of freedom on solid cells. Once we have defined the discrete approximation space, we can solve the problem inside each element KF and KS , communicating the corresponding values at the boundaries ∂KF and ∂KS by the incoming and outgoing fluxes. Thus, the variational problem is represented by a algebraic linear system, with a sparse matrix M, of size equals to the total number f,s f,s of elements Nel multiplied by the number of degrees of freedom per element Ndof , f

f

s that is Nel × Ndof + Nels × Ndof . When compared to the computational cost of standard DG implementation, the corresponding Trefftz-DG cost is thus increased and it is mainly due to the need of inverting the large-sized matrix. The most obvious way to reduce the size of the matrix, which is classically used in most work on space-time Trefftz method, is to consider time slabs. We restrict ourselves to the case of cartesian meshes, but this methodology can also be applied to unstructured meshed. An alternative is to use tent-pitched meshes that respect the causality, this will be the topic of a future work. In order to optimize the execution of the algorithm, we propose to divide the space-time domain Q into Nt elementary time slabs Q1 , Q2 , . . . , QNt and to solve the problem slab by slab, considering the final results, computed in the current time slab at time t, as initial values for the next slab at time t + t (see Fig. 2). Thus the size of matrix inside each time slab is Nt

Fig. 2 Example of 1D + time mesh Th on Q decomposed into Nt time slabs

T = N t Δt

Nt

T = 2Δt

KF

KS t 0 = Δt T = Δt t0 = 0

ΩF

ΩS

Explicit Polynomial Trefftz-DG Method for Space-Time Elasto-Acoustics

401

times smaller, compared to the initial one. Moreover, if the medium parameters are fixed in time, and the space discretization is preserved from slab to slab, the matrix can be computed and inverted once, and then re-projected onto the next time slabs, reducing thus the global numerical cost.

2.2 Polynomial Basis One of the important advantages of Trefftz type methods is the flexibility in the choice of basis functions provided they satisfy the Trefftz property locally in each element. To perform the numerical simulations, we have extended the algorithm proposed by Maciag in [9] for computing wave polynomials, solutions of the second order transient wave equation, to the first order acoustic and elastodynamic systems of dimension one and higher. It consists in computing a polynomial basis, defined in the reference element, using Taylor expansions of generating exponential functions which are local solutions of the initial system of equations. An example of spacetime wave polynomial basis for the first-order acoustic wave equation reads as follows (approximation degree p=3, dimension of the physical space d = 1): φˆ 2v = 1 p φˆ = 0

φˆ 1v = 0 p φˆ = −cF 1

φˆ 5v = − 2 − p φˆ = c2 xt x2

5

F

φˆ 3v = x p φˆ = −c2 t

2

cF2 t 2 2

3

φˆ 6v = −cF xt 2 p φˆ = cF ( x + 6

2

φˆ 7v cF2 t 2 ˆp 2 ) φ7

F

xc2 t 2 = − 6 − 2F 2 c3 t 3 = cF ( F6 + x 2cF t ) x3

φˆ 4v = cF t p φˆ = −cF x 4

3 3

c t φˆ 8v = − F6 − x 2cF t 3 xc2 t 2 p φˆ 8 = cF ( x6 + 2F ) 2

p This basis contains the couples of polynomial functions (φˆ·v , φˆ· ), corresponding to the velocity and pressure respectively, which are locally defined and satisfy the Trefftz property inside each element of the mesh, and of degrees less or equal to p (p = 0, 1, 2, 3) to provide an approximation of order p. By their construction, the Trefftz basis functions are not attached to the coordinates of the degrees of freedom inside the element, contrary to the Lagrange polynomials. Even if we compute only surface integrals, we can evaluate the final approximation solution in any point of the element refinement. We refer to [2] for more numerical details as well as for the acoustic and elastodynamic basis examples of higher dimensions.

2.3 Inversion of the Matrix M Inside a Time Slab The inversion of the matrix inside the time slab can be explicitly reduced to the inversion of its block-diagonal component, which corresponds to the integration at the bottom and top of the time slab (initial and final time faces F0h and FTh ), thanks to the Taylor expansion formulas. More precisely, let us recall the expression for the

402

H. Barucq et al.

bilinear form AT DG (· ; ·) from Sect. 1.2:        . AT DG ·; · ≡ + + + + FTh

F0h

F GH I A T DG

Q

Fh

F

FD h

GH

FFh S

I

AIT DG

It consists of A T DG (·; ·), that corresponds to the integration at the initial and final time element faces of the time slab, and AIT DG (·; ·), that corresponds to the integration at the internal, boundary and fluid-solid element faces. Thus, the matrix M can be represented by the sum of two matrices  M and I MI corresponding I to A T DG and AT DG respectively, as follows: M =  M + I MI . Here,  ∝ ( x)d represents the area of the local faces in F0h and FTh , and I ∝ Q FS ( x)d−1 t represents the area of the local faces in Fh , FD h and Fh respectively. We refer to [2] for more details. This decomposition is of particular interest since M is block-diagonal, each block corresponding to one element. Indeed, we have:    

I −1   M MI =  M I + κP ,

 M + I MI =  M I +



I −1

t Here I is the identity matrix, κ ≡ ∝ x , and P ≡ M MI .  If ||κP || is sufficiently small, we can apply the Maclaurin formula in order to obtain the polynomial expansion for M −1 as follows: ∞ −1    −1  −1  M −1 ≡ I + κP

 M = (−1)n κ n P n  M . n=0

This representation reduces the inversion of the sparse matrix M to the inversion −1 and the multiplication of the inverted blockof its block-diagonal component M diagonal M by the sparse MI . It provides an explicit way for solving the initial linear system approximately. Even though it requires a CFL—type condition related to value of ||κP ||, justifying the approximate solution of the system, it significantly accelerates the algorithm execution. In Table 1 we compare the numerical accuracy (L2 -norm in time and space of numerical error as a function of cell size x) of the TDG method in a 2D homogeneous acoustic case for both the exact and approximate matrix inversions as a function of the mesh size and of the number n of terms in the Taylor expansion.

Explicit Polynomial Trefftz-DG Method for Space-Time Elasto-Acoustics

403

Table 1 Accuracy (L2 -error in space and in time) of the solution when using the approximate inversion with n = 3, 4, 5 and the exact inversion n

x = 10−2

x = 2 · 10−2

x = 5 · 10−2 The approximate inversion (κ = 10−2 ). Accelerating factor ≈18 times 3 1.4166e−05 4.3741e−05 2.8780e−04 4 3.1623e−07 1.2656e−06 5.3868e−05 5 2.8903e−07 9.1744e−07 4.1029e−05 The exact inversion (κ = 10−2 ) · 2.2540e−07 8.9583e−07 5.5811e−05

x = 10−1 2.5772e−03 1.2674e−03 1.3010e−03 1.3004e−03

The accelerating factor is the ratio of the computational costs of the two methods for reaching the same accuracy

3 Numerical Tests For the numerical implementation of the Trefftz-DG method we have considered a 2D medium composed of two homogeneous rectangular layers: the acoustic one and the elastodynamic one. We have set a source term at the fluid-solid interface, and two receivers in the acoustic layer and in the elastodynamic one. The numerical signals at both receivers have been validated with the analytical solutions computed with Gar6more code [5]. In Fig. 3 we show the convergence of the numerical velocity as a function of cell size for different degrees of approximation (p = 0, 1, 2, 3) computed at receivers in (a) 2D acoustic layer and (b) 2D elastodynamic layer. In each case, the convergence rate is higher than the corresponding approximation degree. We refer to [2], where we provide more examples.

(a)

(b)

Fig. 3 Convergence of numerical velocity in function of cell size x. (a) 2D acoustic layer. (b) 2D elastodynamic layer

404

H. Barucq et al.

4 Conclusion The Trefftz-DG methodology for solving the first order elasto-acoustic system has demonstrated the important advantages, such as the use of degrees of freedom evaluated at the element faces only, the flexibility in the choice of the basis functions and the unconditional stability. However, in its initial form, it still shows some limitations due to the space-time integration that leads to the representation of the discrete system by a huge sparse matrix whose straightforward inversion is very expensive, even when using time slabs. We find ourselves in a situation of using an implicit scheme for solving the forward problem that risks to overload the iterative process of the corresponding inverse problem in order to reconstruct very large propagation domains. Fortunately, thanks to the decomposition of the matrix by separating the time variables from the space ones, we could benefit from the block-diagonal structure of the standard DG formulation ending up with an explicit scheme, that is more convenient from the numerical point of view. The performed numerical tests clearly illustrate the interest of the split version of discrete problem. Acknowledgements This project is supported by the Inria—Total SA strategic action “Depth Imaging Partnership” (http://dip.inria.fr), and has received funding from the European Union’s Horizon 2020 research and innovation program under the Marie Sklodowska-Curie grant agreement Number 777778 (Rise action Mathrocks).

References 1. Banjai, L., Georgoulis, E.H., Lijoka, O.: A Trefftz polynomial space-time discontinuous Galerkin method for the second order wave equation. SIAM J. Numer. Anal. 55(1), 63–86 (2017) 2. Barucq, H., Calandra, H., Diaz, J., Shishenina, E.: Space–Time Trefftz - Discontinuous Galerkin Approximation for Elasto-Acoustics. RR-9104, Inria Bordeaux Sud-Ouest, UPPA (LMA-Pau), Total SA, (2017) 3. Barucq, H., Calandra, H., Diaz, J., Shishenina, E.: Space–time Trefftz-DG approximation for elasto-acoustics. Appl. Anal. 1–14 (2018) 4. Egger, H., Kretzschmar, F., Schnepp, S.M., Weiland, T.: A space-time discontinuous Galerkin Trefftz method for time dependent Maxwell’s equations. SIAM J. Sci. Comput. 37(5), B689– B711 (2015) 5. Gar6more2D. Magique-3D (2013). https://gforge.inria.fr/projects/gar6more2d/ 6. Herrera, I.: Trefftz method: a general theory. Numer. Methods Partial Differ. Equ. Int. J. 16(6), 561–580 (2000) 7. Hesthaven, J.S., Warburton, T.: Nodal Discontinuous Galerkin Methods: Algorithms, Analysis, and Applications. Springer Science & Business Media, Berlin (2007) 8. Kretzschmar, F., Moiola, A., Perugia, I., Schnepp, S.M.: A priori error analysis of space–time Trefftz discontinuous Galerkin methods for wave problems. IMA J. Numer. Anal. 36(4), 1599– 1635 (2015) 9. Maciag, A.: The usage of wave polynomials in solving direct and inverse problems for twodimensional wave equation. Int. J. Numer. Methods Biomed. Eng. 27(7), 1107–1125 (2011) 10. Moiola, A., Perugia, I.: A space–time Trefftz discontinuous Galerkin method for the acoustic wave equation in first-order formulation. Numer. Math. 138(2), 389–435 (2018)

Explicit Polynomial Trefftz-DG Method for Space-Time Elasto-Acoustics

405

11. Petersen, S., Farhat, C., Tezaur, R.: A space-time discontinuous Galerkin method for the solution of the wave equation in the time domain. Int. J. Numer. Methods Eng. 78(3), 275– 295 (2009) 12. Trefftz, E.: Ein Gegenstück zum Ritzschen Verfahren. Proceedings of the 2nd International Congress for Applied Mechanics, Zurich, pp. 131–137 (1926) 13. Wang, D., Tezaur, R., Farhat, C.: A hybrid discontinuous in space and time Galerkin method for wave propagation problems. Int. J. Numer. Methods Eng. 99(4), 263–289 (2014)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

An hp-Adaptive Iterative Linearization Discontinuous-Galerkin FEM for Quasilinear Elliptic Boundary Value Problems Paul Houston and Thomas P. Wihler

1 Introduction In this article, we consider the a posteriori error analysis, in a natural meshdependent energy norm, for a class of interior-penalty hp-version discontinuous Galerkin finite element methods (DGFEMs) for the numerical solution of the following quasilinear elliptic boundary value problem: −∇ · (μ(x, |∇u|)∇u) = f

in Ω,

u=0

on Γ.

(1)

Here, Ω ⊂ R2 is a bounded polygon with a Lipschitz continuous boundary Γ , and f ∈ L2 (Ω), where for an open set D ⊆ Ω, we signify by L2 (D) the space of all square integrable functions on D. Additionally, we assume that the nonlinearity μ satisfies the following assumptions: (A1) μ ∈ C0 (Ω × [0, ∞)); (A2) there exist positive constants mμ , Mμ such that mμ (t − s) ≤ μ(x, t)t − μ(x, s)s ≤ Mμ (t − s), ¯ We remark that, if μ satisfies (A2), there exist constants t ≥ s ≥ 0, x ∈ Ω. β ≥ α > 0, such that for all vectors v, w ∈ R2 , and all x ∈ Ω, |μ(x, |v|)v − μ(x, |w|)w| ≤ β|v − w|,   α|v − w|2 ≤ μ(x, |v|)v − μ(x, |w|)w · (v − w);

(2)

P. Houston () School of Mathematical Sciences, University of Nottingham, Nottingham, UK e-mail: [email protected] T. P. Wihler Mathematics Institute, University of Bern, Bern, Switzerland e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_32

407

408

P. Houston and T. P. Wihler

see [14, Lemma 2.1]. For ease of notation, in the sequel, we will simply write μ(s) instead of μ(x, s), thereby suppressing the explicit dependence of μ on x ∈ Ω. The weak formulation of (1) is to find u ∈ H10 (Ω) such that A(u; u, v) = (f, v)L2 (Ω)

∀v ∈ H10 (Ω),

(3)

where, given w ∈ H10 (Ω), we define the bilinear form A(w; u, v) = ; 1 2 Ω μ(|∇w|)∇u ;· ∇v dx, u, v ∈ H0 (Ω), as well as the L (Ω)-inner product 2 1 (v, w)L2 (Ω) = Ω vw dx, v, w ∈ L (Ω). Here, H0 (Ω) is the standard Sobolev space of first order, with zero trace along Γ , equipped with the norm vH1 (Ω) = 0

∇vL2 (Ω) , v ∈ H10 (Ω). Under the assumptions (A1)–(A2) above, it is elementary to show that the form A is strongly monotone and Lipschitz continuous in the sense that A(u; u, u − v) − A(v; v, u − v) ≥ αu − v2H1 (Ω) 0

∀u, v ∈ H10 (Ω),

(4)

and |A(u; u, v) − A(w; w, v)| ≤ βu − wH1 (Ω) vH1 (Ω) 0

0

∀u, v, w ∈ H10 (Ω),

respectively. From these properties, classical monotone operator theory implies existence and uniqueness of a solution of (3); see, e.g., [17, Theorem 3.3.23]. The exploitation of automatic adaptive hp-refinement algorithms has the potential to compute numerical solutions to partial differential equations (PDEs) in a highly efficient manner, often leading to exponential rates of convergence as the underlying finite element space is enriched; see, e.g., [11, 16]. The key tool required to design such strategies is the derivation of a posteriori estimates for the Galerkin discretization errors; in recent years such bounds have been extended to the context of linearization and/or linear solver errors, cf. [1, 2, 4, 5, 7, 9]. In the present article we consider the derivation of an hp-version a posteriori error bound for the DGFEM approximation of the second-order quasilinear elliptic PDE problem stated in (1). To this end, we employ the interior penalty DGFEM proposed in [10], cf. also [12], together with a discrete Kaˇcanov iterative linearization scheme, cf. [6]. Based on the analysis undertaken in [12], together with the use of a suitable reconstruction operator, cf. [13, 15], we derive a fully computable bound for the error, measured in terms of a suitable DGFEM energy norm, which separately accounts for the three main sources of error: discretization, linearization, and linear solver errors. On the basis of this a posteriori bound, we design and implement an hp-adaptive refinement algorithm which automatically controls each of these error contributions as the underlying finite element space is enriched. Numerical experiments highlighting the practical performance of the proposed adaptive strategy are presented.

hp-Adaptive Iterative DGFEMs for Quasilinear PDEs

409

2 Iterative Discontinuous Galerkin Methods 2.1 Discrete hp-Discontinuous Galerkin Spaces Let Th8be a partition of Ω into disjoint open and shape-regular elements κ such that Ω = κ∈Th κ. We assume that each κ ∈ Th is an affine image of a given master element B κ , which is either the open triangle {(x, y) : −1 < x < 1, −1 < y < −x}) or the open square (−1, 1)2 in R2 . By hκ we denote the element diameter of κ ∈ Th , and nκ signifies the unit outward normal vector to κ. We allow Th to be 1-irregular, i.e., each edge of any one element κ ∈ Th contains at most one hanging node (which, for simplicity, we assume to be the midpoint of the corresponding edge). In this context, we suppose that Th is regularly reducible (cf. [18, Section 7.1] and [12]), i.e., there exists a shape-regular conforming (regular) mesh 3 Th (consisting of triangles and parallelograms) such that the closure of each element in Th is a union of closures of elements of 3 Th , and that there exists a constant C > 0, independent of the element sizes, such that for any two elements κ ∈ Th and 3 κ ∈3 Th h κ with 3 κ ⊆ κ we have /h3κ ≤ C. Note that these assumptions imply that Th is of bounded local variation, i.e., there exists a constant ρ1 ≥ 1, independent of the element sizes, such that ρ1−1 ≤ hκ6/hκ7 ≤ ρ1 , for any pair of elements κ6 , κ7 ∈ Th which share a common edge e = ∂κ6 ∩ ∂κ7 . Moreover, let us consider the set E of all one-dimensional open edges of all elements κ ∈ Th . Further, we denote by EI the set of all edges e ∈ E that are contained in the open domain Ω (interior edges). Additionally, we introduce EB to be the set of boundary edges consisting of all e ∈ E that are contained in Γ . For any integer p ∈ N0 , we denote by Pp (κ) the set of polynomials of total degree p on κ. Similarly, when κ is a quadrilateral, we also consider Qp (κ), the set of all tensor-product polynomials on κ of degree p in each coordinate direction. To each κ ∈ Th we assign a polynomial degree pκ (local approximation order). We collect the local polynomial degrees in a vector p = {pκ : κ ∈ Th }, and then introduce the hp-DGFEM space VDG (Th , p) = {v ∈ L2 (Ω) : v|κ ∈ Spκ (κ)

∀κ ∈ Th } ,

with S being either P or Q. We shall suppose that the polynomial degree vector p, with pκ ≥ 1 for each κ ∈ T, has bounded local variation, i.e., there exists a constant ρ2 ≥ 1, independent of the local element sizes and p, such that, for any pair of neighbouring elements κ6 , κ7 ∈ Th , we have ρ2−1 ≤ pκ6/pκ7 ≤ ρ2 . We also define the L2 -projection ΠTh ,p : L2 (Ω) → VDG (Th , p) by (ΠTh ,p v − v, w)L2 (Ω) = 0 ∀w ∈ VDG (Th , p). Evidently, since functions in VDG (Th , p) do not need to be continuous, we have that Πκ,pκ = ΠTh ,p |κ , where, for κ ∈ Th , we let Πκ,pκ be the L2 -projection onto Spκ (κ).

410

P. Houston and T. P. Wihler

2.2 Nonlinear hp-DGFEM Formulation Let κ6 and κ7 be two adjacent elements of Th , and x an arbitrary point on the interior edge e ∈ EI given by e = (∂κ6 ∩ ∂κ7 )◦ . Furthermore, let v and q be scalar- and vector-valued functions, respectively, that are sufficiently smooth inside each element κ6 , κ7 . Then, the averages of v and q at x ∈ e are given by v = 1/2(v|κ6 + v|κ7 ), q = 1/2(q|κ6 + q|κ7 ), respectively. Similarly, the jumps of v and q at x ∈ e are given by [[v]] = v|κ6 nκ6 + v|κ7 nκ7 , [[q]] = q|κ6 · nκ6 + q|κ7 · nκ7 , respectively. On a boundary edge e ∈ EB , we set v = v, q = q and [[v]] = vn, with n denoting the unit outward normal vector on the boundary Γ . Furthermore, we introduce the edge functions h, p ∈ L∞ (E), which, for an edge e ∈ E, are given by h|e := he and p|e = p|e , with he denoting the length of e. In addition, we define the discontinuity penalisation function σ ∈ L∞ (E) given by σ = γ p2 h−1 , where γ ≥ 1 is a (sufficiently large) constant. Then, we equip the DGFEM 2 ; space VDG (Th , p) with the DGFEM norm v2DG := ∇Th v L2 (Ω) + E σ |[[v]]|2 ds, v ∈ VDG (Th , p), where ∇Th is the element-wise gradient operator. With this notation, following [10], we introduce the interior penalty DGFEM discretization of (3) by: find uDG ∈ VDG (Th , p) such that ADG (uDG ; uDG , v) = (f, v)L2 (Ω)

∀v ∈ VDG (Th , p),

(5)

where, for given w ∈ VDG (Th , p), we define the DGFEM bilinear form  ADG (w; u, v) =

Ω

μ(|∇Th w|)∇Th u · ∇Th v dx 



−

E

μ(|∇Th w|)∇Th u · [[v]] ds + θ

 +

EB

σ [[u]] · [[v]] ds,

EB

μ(h−1 |[[w]]|)∇Th v · [[u]] ds

u, v ∈ VDG (Th , p),

where θ ∈ [−1, 1] is a method parameter. Referring to [10, Theorem 2.5], provided that γ ≥ 1 is chosen sufficiently large (independent of the local element sizes and of the polynomial degree distribution), the existence and uniqueness of the DGFEM solution uDG ∈ VDG (Th , p) satisfying (5) is guaranteed. Assumption 1 In the sequel, we suppose that there exists a computable a posteriori error estimate of the form u − uDG DG ≤ η(uDG , f ), where u ∈ H10 (Ω) is the solution of (1), and uDG is its hp-DGFEM approximation defined in (5). Remark 1 In the article [12, Theorem 3.2] it has been proved that such a bound does indeed exist. More precisely, we have that ⎛ u − uDG DG ≤ C ⎝

 κ∈Th

⎞1/2 ηκ2 + O(f, uDG )⎠

=: η(uDG , f ),

(6)

hp-Adaptive Iterative DGFEMs for Quasilinear PDEs

411

where, the local error indicators ηκ , κ ∈ Th , are defined by ηκ2 := h2κ pκ−2 ΠTh ,p−1 (f + ∇ · (μ(|∇uDG )∇uDG ))2L2 (κ) 3 2 + hκ pκ−1 ΠE,p−1 ([[μ(|∇uDG |)∇uDG ]])20,∂κ\Γ + γ 2 h−1 κ pκ [[uDG ]]L2 (∂κ) , (7)

< < (2) and O(f, uDG ) := κ∈Th O(1) κ + e∈EI Oe is a data oscillation term. For κ ∈ Th 2 2 −2 and e ∈ EI , we have O(1) κ := hκ pκ (I−ΠTh ,p−1 )|κ (f +∇·(μ(|∇uDG |)∇uDG ))0,κ , −1 2 and O(2) e := he p¯ e (I − ΠE,p−1 )|e ([[μ(|∇Th uDG |)∇Th uDG ]])0,e , where I denotes a generic identity operator. Here, we write p − 1 := {pκ − 1}κ∈Th . Additionally, we denote by ΠE,p−1 |e the L2 -projector onto Pp¯e −1 (e), where we let p e = max{pκ6 , pκ7 }, with κ6 , κ7 ∈ Th , e = ∂κ6 ∩∂κ7 . Moreover, C > 0 in (6) is a constant that is independent of the local element sizes, the polynomial degree vector p, and the parameters γ and θ .

2.3 Iterative DGFEM In order to provide a practical solution scheme for the nonlinear hp-DGFEM system (5) we propose a linearization approach based on a discrete Kaˇcanov fixed point iteration, see, e.g., [6]. To this end, we begin by selecting an initial guess u0DG ∈ VDG (Th , p). Then, for n ≥ 1, given un−1 ∈ VDG (Th , p), we solve DG the linear hp-DGFEM formulation, defined by ADG (un−1 ; unDG , v) = (f, v)L2 (Ω) DG

∀v ∈ VDG (Th , p),

(8)

for unDG ∈ VDG (Th , p). We emphasize that, in actual computations, the linear system (8) may be solved by an iterative algorithm, thereby generating an approximate numerical solution B unDG ∈ VDG (Th , p), with B unDG ≈ unDG . This means that, in practice, n instead of computing the sequence {uDG }n≥0 obtained from the iteration (8), an inexact sequence {B unDG }n≥0 is generated such that un−1 ;B unDG , v) ≈ (f, v)L2 (Ω) ADG (B DG

∀v ∈ VDG (Th , p).

(9)

From a mathematical view point, this (inexact) iterative linearization DGFEM approach gives rise to three different sources of error: 1. Discretization error, which is expressed by the residual n := ADG (B unDG ; B unDG , ·) − (f, ·)L2 (Ω) . ρDG

(10)

412

P. Houston and T. P. Wihler

n ∈ V (T , p): 2. Linearization error, which is given in terms of the residual ψDG DG h n , v)L2 (Ω) := ADG (B unDG ; B unDG , v) − ADG (B un−1 ;B unDG , v) (ψDG DG

∀v ∈ VDG (Th , p). (11)

n = 0. We observe that, if (1) is linear, then we immediately obtain ψDG n 3. Linear solver error, which is described by a residual λDG ∈ VDG (Th , p):

un−1 ;B unDG , v) − (f, v)L2 (Ω) (λnDG , v)L2 (Ω) := ADG (B DG

∀v ∈ VDG (Th , p). (12)

Note that, if (8) is solved exactly, then we have B un−1 = un−1 and B unDG = unDG , and DG DG n it follows that λDG = 0. Remark 2 Since VDG (Th , p) may not need to be continuous along element intern and λn , respectively, can be faces, the linearization and linear solver residuals ψDG DG computed elementwise, i.e., in parallel, and, hence, at a low computational cost. The aim of the analysis in the following section is to investigate the above residuals, and then to provide a computable a posteriori error estimate for the unDG ∈ VDG (Th , p). error u − B unDG DG between the solution u of (1) and B

2.4 A Posteriori Error Estimation n in (10), we apply an elliptic reconstruction In order to bound the residual ρDG technique along the lines of the works [13, 15], see also [7]. Specifically, we define an auxiliary function 3 un ∈ H10 (Ω) to be the unique solution of the weak formulation n A(3 un ; 3 un , v) = (f + ψDG + λnDG , v)L2 (Ω)

∀v ∈ H10 (Ω),

n and λn are the linearization and linear solver residuals from (11) where ψDG DG and (12), respectively. Upon adding (11) and (12), we notice that n ADG (B unDG ; B unDG , v) = (f + ψDG + λnDG , v)L2 (Ω)

∀v ∈ VDG (Th , p).

un based on In particular, we observe that B unDG is the DGFEM approximation of 3 employing the (nonlinear) DGFEM scheme defined in (5). In particular, we may exploit the a posteriori error estimate in Assumption 1 to infer the computable bound n 3 un − B unDG DG ≤ η(B unDG , f + ψDG + λnDG ).

(13)

hp-Adaptive Iterative DGFEMs for Quasilinear PDEs

413

We now turn to bounding the elliptic reconstruction error u − 3 un ∈ H10 (Ω); to n n this end, we first observe that u − 3 u DG = u − 3 u H1 (Ω) . Then, employing the 0 strong monotonicity property (4), and recalling the weak formulation (3), we obtain αu − 3 un 2DG ≤ A(u; u; u − 3 un ) − A(3 un ; 3 un , u − 3 un ) n ,u −3 un )L2 (Ω) − (λnDG , u − 3 un )L2 (Ω) . = −(ψDG

Employing the Cauchy-Schwarz inequality, together with the Poincaré-Friedrichs inequality, vL2 (Ω) ≤ CPF ∇vL2 (Ω) for all v ∈ H10 (Ω), where CPF > 0 is a constant, we deduce that n u − 3 un DG ≤ ΨDG + ΛnDG ,

(14)

where the linearization and linear solver residuals are given, respectively, by ⎛ ΨDG := CPF/α ⎝ n

 κ∈Th

⎞1/2 ψDG 2L2 (κ) ⎠ n

⎛ ,

ΛDG := CPF/α ⎝ n

 κ∈Th

⎞1/2 λDG 2L2 (κ) ⎠ n

.

Summarizing the above analysis leads to the following result. Theorem 1 Suppose that Assumption 1 is satisfied. Then, given a sequence of (possibly inexact) DGFEM approximations {B unDG }n≥0 ⊂ VDG (Th , p), cf. (9), for n ≥ 1, the following a posteriori error bound holds: n n u − B unDG DG ≤ η(B unDG , f + ψDG + λnDG ) + ΨDG + ΛnDG . n and λn are the residuals defined in (11) Here, u is the analytical solution of (1), ψDG DG and (12), respectively, and α > 0 is the constant occurring in (2) and (4).

Proof The result follows immediately upon application of the triangle inequality, i.e., u − B unDG DG ≤ u − 3 un DG + 3 un − B unDG DG , and inserting the bounds (13) and (14). Remark 3 We note that the above analysis naturally applies to other finite element schemes, provided that Assumption 1 is satisfied.

2.5 Adaptive Iterative hp-DGFEM Procedure In this section we introduce an automatic hp-refinement algorithm which ensures that each of the three components of the error, namely discretization, linearization, and linear solver, are controlled in a suitable fashion. To this end, we propose the following strategy, cf. [9].

414

P. Houston and T. P. Wihler

Algorithm 1 Given a (coarse) starting mesh Th , with an associated (loworder) polynomial degree distribution p, and an initial guess uˆ 0DG ∈ VDG (Th , p). Set n ← 1. 1: Compute the DGFEM solution B unDG from (9) based on employing an iterative linear solver. Furthermore, evaluate the corresponding error indicators n + λn ), Ψ n , and Λn . η(B unDG , f + ψDG DG DG DG 2: if n n + ΛnDG ≤ Υ η(B unDG , f + ψDG + λnDG ) ΨDG

(15)

holds, for some given parameter Υ > 0, then hp-adaptively refine the space VDG (Th , p); go back to step (1:) with the new mesh Th (and based on the previously computed solution B unDG interpolated on the refined mesh). 3: else, i.e., if (15) is not fulfilled, then set n ← n + 1, and perform another linearization step by going back to (1:). 4: end if In Step 2 of Algorithm 1, if (15) is fulfilled then the space VDG (Th , p) is adaptively hp-refined based on first marking elements for refinement according to the size of the local element indicators ηκ , cf. (7). To this end, we exploit the maximal strategy whereby elements are marked for refinement which satisfy the condition ηκ > 1/3 maxκ∈Th ηκ . Secondly, once an element κ ∈ Th has been marked for refinement, we undertake either local mesh subdivision or local polynomial enrichment based on employing the hp-refinement criterion developed within the article [8]. Finally, when (15) is not fulfilled, rather than determining which source n or Λn from (11) and (12), respectively, of error, i.e., the (computable) quantities ΨDG DG is dominant, we choose to always undertake a further linearization step, and hence a further linear solver step is also computed, since this ensures that the most up to date approximation B unDG is employed at all times.

3 Application to Quasilinear Elliptic PDEs In this section we present numerical experiments to highlight the performance of the proposed iterative hp-refinement procedure outlined in Algorithm 1. To this end, we set the interior penalty parameter constant γ to 10 and the steering parameter Υ to 1/4. The solution of the resulting set of linear equations is computed using an ILU(0) preconditioned GMRES algorithm. For the first numerical experiment, we let Ω = (0, 1)2 and define the nonlinear coefficient as μ(|∇u|) = 2 + (1 + |∇u|)−1 . The right-hand forcing function f is selected so that the analytical solution to (1) is given by u(x, y) = x(1 − 2 x)y(1 − y)(1 − 2y)e−20(2x−1) . In Fig. 1 we present a comparison of the actual error measured in terms of the energy norm versus the square root of the number of degrees of freedom in VDG (Th , p). From Fig. 1a we clearly observe exponential convergence of the proposed hp-refinement strategy as VDG (Th , p) in enriched.

hp-Adaptive Iterative DGFEMs for Quasilinear PDEs

415

Furthermore, in Fig. 1b we plot the individual residual error indicators; for this smooth problem, we notice that the discretization indicator (denoted as ηn in the figure) is always dominant, while the linearization and linear solver residuals (denoted as Ψ n and λn , respectively) are roughly of the same magnitude. Secondly, we let Ω denote the L-shaped domain (−1, 1)2 \[0, 1)×(−1, 0] ⊂ R2 and select μ(|∇u|) = 1 + exp(−|∇u|2 ). By writing (r, ϕ) to denote the system of polar coordinates, we choose the forcing function f and an inhomogeneous   boundary condition such that the analytical solution to (1) is u = r 2/3 sin 2/3ϕ , cf. [3]. In Fig. 2 we now present a comparison of the actual error measured in terms of the energy norm versus the third root of the number of degrees of freedom in VDG (Th , p); as before we again attain exponential convergence of the proposed 100

10

True Error Error Estimator

0 n n n

10 -5

10

10-10

10

0

100

200

300

400

-5

-10

0

500

100

200

300

400

500

sqrt(Degrees of Freedom)

sqrt(Degrees of Freedom)

(b)

(a)

Fig. 1 Example 1. (a) Comparison of the DGFEM norm of the error and the a posteriori bound, with respect to the square root of the number of degrees of freedom; (b) individual error estimators

10 0

10

True Error Error Estimator

0 n n

10 -2

10

-2

10

-4

10

-6

n

10 -4 6

8

10

12

14

16

(Degrees of Freedom)1/3

(a)

18

20

10 -8

5

10

15

20

1/3

(Degrees of Freedom)

(b)

Fig. 2 Example 2. (a) Comparison of the DGFEM norm of the error and the a posteriori bound, with respect to the third root of the number of degrees of freedom; (b) individual error estimators

416

P. Houston and T. P. Wihler

hp-refinement strategy as VDG (Th , p) is adaptively refined, though convergence of the a posteriori error estimator is no longer monotonic. Indeed, from Fig. 2b, we observe that once an hp-mesh refinement has been undertaken, then several linearization/solver steps may be required to ensure that the numerical solution has been computed to a sufficient accuracy before future refinements may be undertaken.

4 Conclusions In this article we have derived a computable hp-version a posteriori error bound for the DGFEM approximation of a second-order quasilinear elliptic PDE problem, whereby a discrete Kaˇcanov iterative linearization scheme is employed. The resulting computable upper bound directly takes into account discretization error, as well as the errors stemming from linearization and the underlying linear solver. Numerical experiments highlighting the performance of this bound within an automatic hp-refinement algorithm are presented. Acknowledgements TW acknowledges the support of the Swiss National Science Foundation (SNF), Grant No. 200021_162990.

References 1. Amrein, M., Wihler, T.P.: Fully adaptive Newton-Galerkin methods for semilinear elliptic partial differential equations. SIAM J. Sci. Comput. 37(4), A1637–A1657 (2015) 2. Amrein, M., Melenk, J.M., Wihler, T.P.: An hp-adaptive Newton-Galerkin finite element procedure for semilinear boundary value problems. Math. Methods Appl. Sci. 40(6), 1973– 1985 (2017) 3. Congreve, S., Houston, P., Wihler, T.P.: Two-grid hp-version discontinuous Galerkin finite element methods for second-order quasilinear elliptic PDEs. J. Sci. Comput. 55(2), 471–497 (2013) 4. El Alaoui, L., Ern, A., Vohralík, M.: Guaranteed and robust a posteriori error estimates and balancing discretization and linearization errors for monotone nonlinear problems. Comput. Methods Appl. Mech. Eng. 200(37–40), 2782–2795 (2011) 5. Ern, A., Vohralík, M.: Adaptive inexact Newton methods with a posteriori stopping criteria for nonlinear diffusion PDEs. SIAM J. Sci. Comput. 35(4), A1761–A1791 (2013) 6. Garau, E.M., Morin, P., Zuppa, C.: Convergence of an adaptive Kaˇcanov FEM for quasi-linear problems. Appl. Numer. Math. 61(4), 512–529 (2011) 7. Heid, P., Wihler, T.P.: Adaptive iterative linearization Galerkin methods for nonlinear problems (2018). Technical Report 1808.04990. arxiv.org 8. Houston, P., Süli, E.: A note on the design of hp-adaptive finite element methods for elliptic partial differential equations. Comput. Methods Appl. Mech. Eng. 194(2–5), 229–243 (2005) 9. Houston, P., Wihler, T.P.: An hp-adaptive Newton-discontinuous-Galerkin finite element approach for semilinear elliptic boundary value problems. Math. Comp. 87(314), 2641–2674 (2018)

hp-Adaptive Iterative DGFEMs for Quasilinear PDEs

417

10. Houston, P., Robson, J., Süli, E.: Discontinuous Galerkin finite element approximation of quasilinear elliptic boundary value problems I: the scalar case. IMA J. Numer. Anal. 25, 726– 749 (2005) 11. Houston, P., Schötzau, D., Wihler, T.P.: Energy norm a posteriori error estimation of hpadaptive discontinuous Galerkin methods for elliptic problems. Math. Models Methods Appl. Sci. 17(1), 33–62 (2007) 12. Houston, P., Süli, E., Wihler, T.P.: A posteriori error analysis of hp-version discontinuous Galerkin finite-element methods for second-order quasi-linear elliptic PDEs. IMA J. Numer. Anal. 28(2), 245–273 (2008) 13. Lakkis, O., Makridakis, C.: Elliptic reconstruction and a posteriori error estimates for fully discrete linear parabolic problems. Math. Comp. 75(256), 1627–1658 (2006) 14. Liu, W.B., Barrett, J.W.: Quasi-norm error bounds for the finite element approximation of some degenerate quasilinear elliptic equations and variational inequalities. RAIRO Modél. Math. Anal. Numér. 28(6), 725–744 (1994) 15. Makridakis, C., Nochetto, R.H.: Elliptic reconstruction and a posteriori error estimates for parabolic problems. SIAM J. Numer. Anal. 41(4), 1585–1594 (2003) 16. Melenk, J.M., Wohlmuth, B.I.: On residual-based a posteriori error estimation in hp-FEM. Adv. Comput. Math. 15(1–4), 311–331 (2001). A posteriori error estimation and adaptive computational methods 17. Neˇcas, J.: Introduction to the Theory of Nonlinear Elliptic Equations. John Wiley and Sons, Hoboken (1986) 18. Ortner, C., Süli, E.: Discontinuous Galerkin finite element approximation of nonlinear secondorder elliptic and hyperbolic systems. SIAM J. Numer. Anal. 45(4), 1370–1397 (2007)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Erosion Wear Evaluation Using Nektar++ Manuel F. Mejía, Douglas Serson, Rodrigo C. Moura, Bruno S. Carmo, Jorge Escobar-Vargas, and Andrés González-Mancera

1 Introduction Wear is a common phenomenon on many machines and devices, it is characterised by the removal or loss of material. Erosion wear is a particular wear process which occurs when solid particles or droplets, carried by a fluid (liquid or gas), impact on a solid surface [1]. Turbomachinery such as pumps, turbines and pipe accessories (i.e. tees, elbows, nozzles, valves), are examples of elements affected by the erosion wear, decreasing the performance and the lifetime. In many industrial sectors e.g. energy and mining, and oil & gas; massive amounts of resources are used for maintenance and replacement of affected parts [2–4]. Despite this phenomenon have been broadly investigated [5–14] there are still unsolved challenges in establishing the influence of small eddies during the erosion process leading to modest accuracy levels in the simulation results.

M. F. Mejía () Universidad de Los Andes, Bogotá, D.C., Colombia Universidad Central, Bogotá, D.C., Colombia e-mail: [email protected] D. Serson · B. S. Carmo Universidade de São Paulo (USP), São Paulo, Brazil R. C. Moura Instituto Tecnológico de Aeronáutica, São Paulo, Brazil J. Escobar-Vargas Pontificia Universidad Javeriana, Bogotá, D.C., Colombia A. González-Mancera Universidad de Los Andes, Bogotá, D.C., Colombia © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_33

419

420

M. F. Mejía et al.

Due to the microscopic nature of erosion, the smallest scales in the flow play a fundamental role in the complete process. One of the aspects which has not been carefully studied in erosion wear modelling is the effect that the smaller eddies and secondary flows have on the particles interactions with the surface. In general, these secondary flows could not be represented using linear Reynolds Average Navier Stokes (RANS) simulations, this is mainly because the Reynolds stress imbalance is neglected and the secondary flow does not develop. As was mention by Gross and Fasel [15], predictions of the secondary flow require non-linear Reynolds stress, full Reynolds-stress models, Large Eddy Simulations (LES) or Direct Numerical Simulation (DNS). Due to their relatively low computational cost, RANS models often used to predict on erosion using CFD in industrial simulations. The inclusion of smaller eddies and secondary flows in the simulation could be a major breakthrough in the modelling of erosion process. In order to capture in an accurate way the physics related with the small eddies and secondary flows, a numerical technique capable to represent those processes, is needed. As emphasised by Jacobs [16], the use of spectral methods could allow increased accuracy in the simulation due to the potential to simulate a wider range of scales. With this in mind, the purpose of this work is to assess the impact of higher resolution methods on the prediction of erosion wear rate and distribution.

1.1 Spectral Methods Several numerical techniques are used to solve Navier Stokes (NS) equations. Some of them are finite differences, finite volumes and finite elements. Nevertheless, when high accuracy is required the use of a lot of elements is needed in the modelling, which significantly increase the computational cost [17, 18]. Hence novel methods are subject of research to offer a better rate accuracy and computational cost. Among novel numerical methods considered nowadays are spectral methods, which have shown to be a powerful tool with high level of accuracy for solving large problems in computational fluid dynamics (CFD), according to the available literature, especially in the studies developed by Boyd [19], Canuto et al. [20–22], Trefethen [23, 24] and Sherwin [25, 26]. Nektar++ is an open-source software framework designed to support the development of high-performance scalable solvers for partial differential equations using the spectral/hp element method[27]. High Order CFD methods have been receiving considerable attention in the past two decades. Traditional CFD software could be replaced by high order code in many applications in few years [28].

1.2 Particles Tracking To the best of the authors’ knowledge, there is no work that uses high order methods to evaluate erosion wear rate. This research aims to assess the impact of higher resolution methods on the prediction of erosion wear rate and distribution. It

Erosion Wear Evaluation Using Nektar++

421

comprises the solution of fluid flow using incompressible NS solver with implicit LES modelling, the implementation of a Lagrangian particle tracking model and the later data processing through traditional erosion rate models but using the available high order information. This could allow the evaluation of traditional rate models with more spatial resolution and accuracy. The Lagrangian particle tracking model is based on one-way coupling approach, that is the most simple case when just the iteration between the fluid and each particle is taking into account in just one way. That means that the particles are moved by the fluid but the fluid flow is not perturbed by the particles. Moreover, the effects of the collision between particles are also neglected. The one-way coupling model is valid for volume concentrations of particles lower than 10−6 [29, 30]. The problem of predicting particle motion in a fluid flow can be predicted by solving an evolution equation in time: d vp = F (u, ρ, ρp , Cd , . . .) dt

;

d xp = vp dt

(1)

where vp and xp are the particle velocity and position and F is a function of the velocity of the fluid u, the density of the fluid ρ and particle density ρp , among others. To start, it is necessary to obtain the velocity on a certain point from the eulerian velocity field. This process consists of finding the element containing the particle and interpolating the velocity with the element information. In a higher-order velocity element field, the use of linear interpolation is inaccurate and could vanish the advantage won with the use of high order methods. On the other hand, using high order interpolation could be computationally expensive. Therefore, special attention to this procedure is required [16, 31, 32].

1.3 Erosion Wear Evaluation Once the information about the collisions is complete, the erosion wear model is used to predict the pattern of material removed. The general erosion equation, based on the work of Finnie [33–38] can be presented as W = kFs Vpn f (θ )

(2)

W is the erosion rate or material removed by collision, k is a wall material dependent constant, Fs is the particle geometric factor, Vp and n are the collision velocity and the velocity exponent, and f (θ ) is a function of collision angle. Several authors define these values for different materials configurations and test cases. Three of the most used models, which include experimental results are the jet impingement test [39–45], elbow erosion [46–49], and the works of the Wong et al. [4, 46, 50, 51].

422

M. F. Mejía et al.

2 Implementation This section describes the implementation of erosion wear in Nektar++. To achieve this objective is important to have in mind the partition of the problems into two parts. The first one is the particle tracking as a filter within the Nektar++ incompressible Navier Stokes solver. A filter in Nektar++ is a module for calculating a variety of useful quantities from the field variables as the solution evolves in time [27]. The second one is implemented as a FieldConvert module to evaluate the erosion of each collision and generate the fields on the boundaries walls. FieldConvert is a utility embedded in Nektar++ with the primary aim of allowing the user to work with the Nektar++ output files, some of the modules within FieldConvert allow the user to postprocess the output data [27].

2.1 Particles Tracking The first step was the implementation of a ODE time solver. Several options are available, but having into account the discrete time flow fields calculated with the Navier Stokes incompressible solver, and to avoid the use of temporal interpolation, the selected option was the Adams-Bashforth (AB) and Adams-Moulton (AM) schemes. The implementation was tested with a benchmark case presented in [31]. In this model, the particle velocity is the fluid velocity at certain point and the evolution equation is reduced to one equation; Eq. 1 is reduced to: d xp =u dt

(3)

To solve this system a Time-Marching Method was implemented, meaning that the future values are evaluated using the present and past values of the variables. Explicit AB and Implicit AM methods were implemented using first to fourth integration order. The error values obtaining using AB and AM with different order presents features from this kind of methods. The next step was the implementation of the solid particles. In this case, the momentum equation is evaluated on each particle, resulting:   d vp ρp − ρ = Fd u − vp + g dt ρp Fd =

3 Cd Rep ρp 4 νdp2

;

d xp = vp dt

(4) (5)

Erosion Wear Evaluation Using Nektar++ u−v d Rep = ( νp ) p

⎧ ⎪ ⎪ Rep < 0.5 ⎪ ⎨24/Rep ,  0.687 , 0.5 < Rep < 1000 Cd = 24/Rep 1 + 0.15Rep ⎪ ⎪ ⎪ ⎩0.44, Rep > 1000

423

(6)

(7)

where Fd is the drag force, Rep is the Reynolds Number based on the diameter of the particle, g is the gravity acceleration, and Cd is the drag force evaluated on each particle. Figure 1 shows a diagram of the evolution equation. Current position, velocities and forces are evaluated to get the future positions (BP, OP) until the next position is located outside of the domain (NP). When this happens, the evolution algorithm stop and a function is used to evaluate the collision point (CP) and the position after of collision (NP’) using the high order information about the walls.

2.2 Erosion Wear Evaluation Erosion rate per collision (Eq. 2) has to be integrated over each element of the eroded surface. For each particle collision, more material is removed from the surface, the elemental erosion rate has to take into account this cumulative effect over the surface. As mentioned before, the set of parameters used in this work, has been based on experimental data. One of the most used parameter set is the one proposed by Erosion group of the University of Tulsa [38, 48, 52]. The erosion rate takes the form of Eq. 2, Fs = 1 for sharp (angular), 0.53 for semi-rounded, or 0.2 for fully rounded sand particles. Vp is the impact velocity and n = 1.73. The angle function has the form: ⎧ ⎨aθ 2 + bθ, for θ ≤ φ (8) f (θ ) = ⎩x cos2 (θ ) sin(wθ ) + y sin2 (θ ) + z for θ > φ

Fig. 1 Evolution of particle tracking

424

M. F. Mejía et al.

All the parameters and empirical constants depend on the material being eroded. For velocity in ft/s, the steel-sand parameters are: a = 38.4, b = 22.7 φ = 1, x = 0.3147, y = 0.03609, w = 0.2532 and z = 0 [53].

3 Test Case To test the new feature in Nektar++, a Backward Facing Step (BFS) model was developed based on the experimental setup of [30, 32, 54] showed in Fig. 2. In the model developed in this work, the simulations were done with the addition of gravitational effects on the −y direction. In original experiments the air at the inlet is a well development turbulent flow (u¯ = 10.5 m/s), this is used a inlet condition and, to complete the model, a zero pressure condition at the output. The additional boundaries were set as walls. A zero velocity field was set as initial condition. The particles used have a 70 µm diameter and 8808 kg/m3 density. Figure 3 (top) shows a snapshot of the velocity field when the statistically stationary regime is reached (t = 8 s), next the particles are released and were convected by the flow. Particular trajectories are shown in grey lines in Fig. 3 (bottom). In the same figure, results of the particle collision with the walls, computed with Eq. 2, are also shown. From the results presented Fig. 3, the typical BFS velocity profiles can be recognised. It is important to note the details behind the step, the main flow originates the secondary eddies and defines the limit of the recirculation zone (x/H = 7 from the step) where backflow occurs. Additionally, interesting details appear in between each main velocity flow ripple and the walls along the x-direction. It is noteworthy that particle tracking is evaluated using a steady velocity field, therefore the existence of several irregularities is expected, for instance, particle trajectories inside recirculation zone. Erosion rate depends on the number of collisions at specific points. It is a localised phenomenon that does not occur continuously in the domain. Its distribution shows a strong dependence on the flow dynamics.

Fig. 2 Geometry of the Backward Facing Step setup. The initial velocity was set to get a Re = 18,600

Erosion Wear Evaluation Using Nektar++

425

Fig. 3 BFS case results. Top: Velocity field at a statistically stationary condition. Bottom: Distribution of the particles inside the flow (gray lines). The colours in the walls indicate the location of the normalised erosion rate

4 Conclusion and Future Work This work presented a method developed to asses the erosion wear rate using a high-order (spectral) element based technique on a modified test case implemented in Nektar++. The methodology proposed in this study have a potential to increase the accuracy when solving this kind of problems. Future research activities are going to be focused on the determination of accuracy improvements and optimisation of the proposed methodology. Several more cases have to be tested to produce solid conclusions about the implemented methodology, as well as a detailed comparison with experimental test cases. Despite the methodology implemented had several important simplifications, as the use of one-way coupling and the few forces taken into account, allowed quicker implementations and results. This work would be an interesting starting to implement this kind of simulations using Nektar++. However, to run more realistic cases, additional research efforts are required for the implementation of two-way and four-way coupling and the effects of other forces over each particle.

References 1. Kulu, P., Kleis, I.: Solid Particle Erosion: Occurrence, Prediction and Control. Springer, London (2008) 2. Neopane, H., Cervantes, M.: Sediment erosion in hydraulic turbines. Global J. Res. Eng. Mech. Mech. Eng. 11, 17–26 (2011)

426

M. F. Mejía et al.

3. Walker, C.I., Robbie, P.: Comparison of some laboratory wear tests and field wear in slurry pumps. Wear 302(1–2), 1026–1034 (2013) 4. Wong, C.Y., Solnordal, C., Graham, L., Short, G., Wu, J.: Slurry erosion of surface imperfections in pipeline systems. Wear 336, 72–85 (2015) 5. Karimi, A., Schmid, R.K.: Ripple formation in solid-liquid erosion. Wear 156(1), 33–47 (1992) 6. Zeng, C., Xiao, Y., Wang, Z., Zhang, J., Luo, Y.: Numerical analysis of a Pelton bucket free surface sheet flow and dynamic performance affected by operating head. Proc. IMechE 231(3), 182–196 (2017) 7. Padhy, M.K., Saini, R.P.: Study of silt erosion mechanism in Pelton turbine buckets. Energy 39(1), 286–293 (2012) 8. Bajracharya, T.R., Acharya, B., Joshi, C.B., Saini, R.P., Dahlhaug, O.G.: Sand erosion of Pelton turbine nozzles and buckets: a case study of Chilime Hydropower Plant. Wear 264(3–4), 177– 184 (2008) 9. Kumar, P., Saini, R.P.: Study of cavitation in hydro turbines—a review. Renew. Sust. Energ. Rev. 14(1), 374–383 (2010) 10. Xiao, Y.-X., Han, F.-Q., Zhou, J.-L., Kubota, T.: Numerical prediction of dynamic performance of Pelton turbine. J. Hydrodyn. B 19(3), 356–364 (2007) 11. Zhu, H., Pan, Q., Zhang, W., Feng, G., Li, X.: CFD simulations of flow erosion and flowinduced deformation of needle valve: effects of operation, structure and fluid parameters. Nucl. Eng. Des. 273, 396–411 (2014) 12. Chongji, Z., Yexiang, X., Wei, Z., Yangyang, Y., Lei, C., Zhengwei, W.: Pelton turbine needle erosion prediction based on 3D three- phase flow simulation. IOP Conf. Ser. Earth Environ. Sci. 22(5), 052019 (2014) 13. Finnie, I., Kabil, Y.H.: On the formation of surface ripples during erosion. Wear 8(1), 60–69 (1965) 14. Dutta, S.: Bulle-effect and its implications for morphodynamics of river diversions. Ph.D. Thesis. University of Illinois at Urbana-Champaign (2017) 15. Gross, A., Fasel, H.: Hybrid RANS/LES simulations of turbulent channel and diffuser flows. In: 48th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition, Aerospace Sciences Meetings. American Institute of Aeronautics and Astronautics (2010) 16. Jacobs, G.: Inertial particle behavior in an unsteady separated flow. In: 46th AIAA Aerospace Sciences Meeting and Exhibit, Aerospace Sciences Meetings. American Institute of Aeronautics and Astronautics (2008) 17. Sengupta, K., Mashayek, F., Jacobs, B.: Direct numerical simulation of turbulent flows using spectral methods. In: 46th AIAA Aerospace Sciences Meeting and Exhibit, pp. 7–10 (2008) 18. Olige, J., Kreiss, H.O.: Comparison of accurate methods for integration of hyperbolic equations. Tellus 24, 199–215 (1972) 19. Boyd, J.P.: Chebyshev and Fourier Spectral Methods, 2nd edn. Dover Publications, Mineola (2000) 20. Canuto, C., Hussaini, M.Y., Quarteroni, A., Zang, T.A.: Spectral Methods: Fundamentals in Single Domains. Scientific Computation. Springer, Berlin (2007) 21. Canuto, C, Hussaini, M.Y., Quarteroni, A., Zang, T.A.: Spectral Methods in Fluid Dynamics. Springer, New York (1987) 22. Zang, T.A., Canuto, C., Hussaini, M.Y., Quarteroni, A.: Spectral Methods: Complex Geometries and Applications to Fluid Dynamics, 1st edn. Springer, Berlin (2007) 23. Trefethen, L.N.: Finite difference and spectral methods for ordinary and partial differential equations, 1st edn. Cornell University, Ithaca (1996) 24. Trefethen, L.N.: Spectral Methods in Matlab, 1st edn. Cornell University, Ithaca (2000) 25. Karniadakis, G.E.M., Sherwin, S.: Spectral/hp Element Methods for Computational Fluid Dynamics. Oxford University Press, New York (2005) 26. Karniadakis, G.E., Sherwin, S.J.: Spectral/hp Element Methods for CFD. Oxford University Press, New York (1999)

Erosion Wear Evaluation Using Nektar++

427

27. Cantwell, C.D., Moxey, D., Comerford, A., Bolis, A., Rocco, G., Mengaldo, G., De Grazia, D., Yakovlev, S., Lombard, J.-E., Ekelschot, D., Jordi, B., Xu, H., Mohamied, Y., Eskilsson, C., Nelson, B., Vos, P., Biotto, C., Kirby, R.M., Sherwin, S.J.: Nektar++: an open-source spectral/ element framework. Comput. Phys. Commun. 192, 205–219 (2015) 28. Wang, Z.J., Fidkowski, K., Abgrall, R., Bassi, F., Caraeni, D., Cary, A., Deconinck, H., Hartmann, R., Hillewaert, K., Huynh, H.T., Kroll, N., May, G., Persson, P.-O., van Leer, B., Visbal, M.: High-order CFD methods: current status and perspective. Int. J. Numer. Methods Fluids 72(8), 811–845 (2013) 29. Elghobashi, S: On predicting particle-laden turbulent flows. Appl. Sci. Res. 52(4), 309–329 (1994) 30. Greifzu, F., Kratzsch, C., Forgber, T., Lindner, F., Schwarze, R.: Assessment of particletracking models for dispersed particle-laden flows implemented in OpenFOAM and ANSYS FLUENT. Eng. Appl. Comput. Fluid Mech. 10(1), 30–43 (2016) 31. Coppola, G., Sherwin, S.J., Peiro, J.: Nonlinear particle tracking for high-order elements. J. Comput. Phys. 172, 356–386 (2001) 32. Jacobs, G.B., Kopriva, D.A., Mashayek, F.: Towards efficient tracking of inertial particles with high-order multidomain methods. J. Comput. Appl. Math. 206(1), 392–408 (2007) 33. Finnie, I.: Erosion of surfaces by solid particles. Wear 3(2), 87–103 (1960) 34. Finnie, I.: Some observations on the erosion of ductile metals. Wear 19(1), 81–90 (1972) 35. Padhy, M.K., Saini, R.P.: A review on silt erosion in hydro turbines. Renew. Sust. Energ. Rev. 12(7), 1974–1987 (2008) 36. Finnie, I.: Some reflections on the past and future of erosion. Wear 186–187, 1–10 (1995) 37. Hutchings, I., Shipway, P.: Tribology: Friction and Wear of Engineering Materials. Butterworth-Heinemann, London (2017) 38. Parsi, M., Najmi, K., Najafifard, F., Hassani, S., McLaury, B.S., Shirazi, S.A.: A comprehensive review of solid particle erosion modeling for oil and gas wells and pipelines applications. J. Nat. Gas Sci. Eng. 21, 850–873 (2014) 39. Sugiyama, K., Harada, K., Hattori, S.: Influence of impact angle of solid particles on erosion by slurry jet. Wear 265(5), 713–720 (2008) 40. Gnanavelu, A., Kapur, N., Neville, A., Flores, J.F., Ghorbani, N.: A numerical investigation of a geometry independent integrated method to predict erosion rates in slurry erosion. Wear 271(5), 712–719 (2011) 41. Efird, K.D., Wright, E.J., Boros, J.A., Hailey, T.G.: Correlation of steel corrosion in pipe flow with jet impingement and rotating cylinder tests. Corrosion 49(12), 992–1003 (1993) 42. López, A., Nicholls, W., Stickland, M.T., Dempster, W.M.: CFD study of jet impingement test R and OpenFOAM6. R Comput. Phys. Commun. 197, 88–95 (2015) erosion using Ansys Fluent6 43. Gnanavelu, A., Kapur, N., Neville, A., Flores, J.F.: An integrated methodology for predicting material wear rates due to erosion. Wear 267(11), 1935–1944 (2009) 44. Sugiyama, K., Nakahama, S., Hattori, S., Nakano, K.: Slurry wear and cavitation erosion of thermal-sprayed cermets. Wear 258(5–6), 768–775 (2005) 45. Oka, Y.I., Hayashi, H.: Evaluation of erosion resistance for metal–ceramic composites and cermets using a water-jet testing apparatus. Wear 271(9–10), 1397–1403 (2011) 46. Solnordal, C.B., Wong, C.Y., Boulanger, J.: An experimental and numerical analysis of erosion caused by sand pneumatically conveyed through a standard pipe elbow. Wear 336–337, 43–57 (2015) 47. Asgharpour, A., Zahedi, P., Khanouki, H.A., Shirazi, S.A., McLaury, B.S.: Experimental and numerical study on solid particle erosion in elbows mounted in series. In: ASME 2017 Fluids Engineering Division Summer Meeting, p. V01BT06A015. American Society of Mechanical Engineers, New York (2017) 48. Shirazi, S.A., Shadley, J.R., McLaury, B.S., Rybicki, E.F.: A procedure to predict solid particle erosion in elbows and tees. J. Press. Vessel. Technol. 117(1), 45–52 (1995) 49. Vieira, R.E., Mansouri, A., McLaury, B.S., Shirazi, S.A.: Experimental and computational study of erosion in elbows due to sand particles in air flow. Powder Technol. 288, 339–353 (2016)

428

M. F. Mejía et al.

50. Wong, C.Y., Solnordal, C., Swallow, A., Wu, J.: Experimental and computational modelling of solid particle erosion in a pipe annular cavity. Wear 303(1), 109–129 (2013) 51. Boulanger, J., Wong, C., Solnordal, C., Zamberi, MSA., Shaffee, A., Johar, Z., Jadid, M.: A simplified computational approach to multi-phase erosion. International Conference on CFD in the Minerals and Process Industries, CSIRO, vol. 11 (2015) 52. Chen, X., McLaury, B.S., Shirazi, S.A.: Application and experimental validation of a computational fluid dynamics (CFD)-based erosion prediction model in elbows and plugged tees. Comput. Fluids 33(10), 1251–1272 (2004) 53. Ahlert, K.R.: Effects of particle impingement angle and surface wetting on solid particle erosion of AISI 1018 steel (1994) 54. Fessler, J.R., Eaton, J.K.: Turbulence modification by particles in a backward-facing step flow. J. Fluid Mech. 394, 97–117 (1999)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

An Inexact Petrov-Galerkin Approximation for Gas Transport in Pipeline Networks Herbert Egger, Thomas Kugler, and Vsevolod Shashkov

1 Introduction The flow of gas in a horizontal pipeline of constant cross section is described by [2] A∂t ρ + ∂x m = 0 $ # λ |m| m2 + Ap = − m. ∂t m + ∂x Aρ 2D Aρ

(1) (2)

Here A and D are the cross section and diameter of the pipe, and λ is a dimensionless friction parameter. The functions ρ, p, and m describe the density, pressure, and mass flow rate of the gas. Under isothermal flow conditions, one has p = c2 ρ

(3)

with constant c denoting the speed of sound. In practically relevant scaling regimes, the nonlinear term on the left hand side of (2) is usually neglected, which can be justified by an asymptotic analysis [2, 7]. Using this simplification and Eq. (3) to

H. Egger () · T. Kugler TU Darmstadt, Department of Mathematics, Darmstadt, Germany e-mail: [email protected]; [email protected] V. Shashkov TU Darmstadt, GSC Computational Engineering, Darmstadt, Germany e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_34

429

430

H. Egger et al.

eliminate the density, one arrives at evolution problems of the general form a∂t p + ∂x m = 0

(4)

b∂t m + ∂x p = −dm

(5)

where a and b are positive constants and d = d(p, m) denotes a state dependent friction coefficient. For our analysis, we will consider d = d(x) as a function depending only on space which can be justified, e.g., by linearization around a steady state. Corresponding models for the gas flow on pipe networks are obtained by coupling the flow equations for single pipes via algebraic conditions [9, 10]; see below. The discretization of (4)–(5) and its extension to pipeline networks has been discussed intensively in the literature. In [9], a Galerkin approximation for (1)–(2) with cubic Hermite polynomials is investigated numerically. The discretization of transient gas flow models is also studied [2, 5, 8]. An entropy stable finite volume method is proposed in [10], and an energy stable mixed finite element approximation is investigated in [3]. Apart from [9], all methods discussed above are of lowest order and no rigorous convergence analysis is given. In this paper, we study the discretization of (4)–(5) by a Petrov-Galerkin approach of potentially high order. The resulting scheme is shown to be stable which allows us to prove order optimal convergence rates. By using an appropriate functional analytic setting, the convergence results can be generalized almost verbatim to pipeline networks. A hybridization strategy will be discussed that facilitates the implementation and that allows to incorporate non-standard coupling conditions. The proposed method formally also allow to treat nonlinear models of gas transport and, in principle, high order convergence can be obtained in practically relevant regimes.

2 Notation and Preliminaries Let xL < xR and denote by Lp (xL , xR ) and W k,p (xL , xR ), k ≥ 0 the standard Lebesgue and Sobolev spaces. The scalar product and norm of L2 (xL , xR ) are written as (v, w) and v = vL2 . Other norms will be designated by subscripts. We write H k (xL , xR ) = W k,2 (xL , xR ) for the Hilbert spaces and define H01 = {v ∈ H 1 (xL , xR ) : v(xL ) = v(xR ) = 0} and H (div) = H 1 (xL , xR ) for convenience. The reason for introducing the space H (div) will become clear when considering networks, where the spaces H 1 and H (div) have different continuity properties across junctions. By Lp (0, T ; X) and W k,p (0, T ; X) we denote the Bochner spaces of functions f : [0, T ] → X with values in X. The value of f (t) may then itself be a function. In the following, we consider the linear

An Inexact Petrov-Galerkin Approximation for Gas Transport in Pipeline Networks

431

system a∂t p(x, t) + ∂x m(x, t) = f (x, t),

(6)

b∂t m(x, t) + ∂x p(x, t) + d(x)m(x, t) = g(x, t),

(7)

for xL < x < xR and t > 0 with homogeneous boundary conditions p(xL , t) = p(xR , t) = 0.

(8)

Inhomogeneous and more general boundary conditions can be considered as well and our analysis applies with minor modifications. We will assume that (A1) (A2)

a, b are positive constants, and d ∈ L∞ (xL , xR ) with 0 < d ≤ d(x) ≤ d and constants d, d.

For given f, g ∈ L2 (0, T ; L2 (xL , xR )) and initial values p(0) ∈ H01 , m(0) ∈ H (div), existence of a unique solution follows from semigroup theory. Any smooth solution of problem (6)–(8) also satisfies p(t) ∈ H01 , m(t) ∈ H (div), and (a∂t p(t), 3 q ) + (∂x m(t), 3 q ) = (f (t), 3 q)

(9)

(b∂t m(t),3 v ) + (∂x p(t),3 v ) + (dm(t),3 v ) = (g(t),3 v)

(10)

for all 3 v, 3 q ∈ L2 (xL , xR ) and all 0 < t < T . This variational characterization will be the starting point for our discretization approach introduced in the next section.

3 Petrov-Galerkin Approximation Let xL = x0 < x1 < . . . < xN = xR be a partition of the interval [xL , xR ] into elements Tn = [xn−1 , xn ]. We call Th := {Tn : 1 ≤ n ≤ N} the mesh and denote by hn = |xn − xn−1 | and h = maxn hn the local and global mesh size, respectively. Let Pk (Th ) := {v ∈ L2 (xL , xR ) : v|T ∈ Pk (T ) ∀T ∈ Th }

(11)

be the space of piecewise polynomials on the mesh Th . We fix k ≥ 1 and search for approximations for the solutions p(t), m(t) of problem (6)–(8) in the spaces Qh = Pk (Th ) ∩ H01

and Vh = Pk (Th ) ∩ H (div)

(12)

of continuous piecewise polynomials with appropriate boundary conditions. As finite dimensional test spaces for the variational problem (9)–(10), we choose 3h = Pk−1 (Th ) Q

3h = Pk−1 (Th ) and V

(13)

432

H. Egger et al.

consisting of discontinuous piecewise polynomials of lower order k − 1. We denote by Ihk : H 1 (xL , xR ) → Pk (Th ) ∩ H 1 (xL , xR ) the H 1 -projection operator, defined by (Ihk v)(xk ) = v(xk ) and

vh ) = (∂x v,3 vh ) (∂x Ihk v,3

for all 0 ≤ k ≤ N,

(14)

for all 3 vh ∈ Pk−1 (Th ),

(15)

and let πhk−1 : L2 (xL , xR ) → Pk−1 (Th ) be the L2 -orthogonal projection, satisfying (πhk−1 v,3 vh ) = (v,3 vh )

for all 3 vh ∈ Pk−1 (Th ).

(16)

Note that both projection operators Ihk and πhk−1 can be defined locally on every element. Moreover, they are mutually related to each other by the commuting diagram property ∂x Ihk v = πhk−1 ∂x v

for all v ∈ H 1 (xL , xR ).

(17)

For the approximation of problem (6)–(8), we then use the following approximation. Problem 1 (Inexact Petrov-Galerkin Method) Find functions ph ∈ H01 (0,T ;Qh ), mh ∈ H 1 (0, T ; Vh ) with ph (0) = Ihk p(0) and mh (0) = Ihk m(0), and such that (a∂t ph (t), 3 qh ) + (∂x mh (t), 3 qh ) = (f (t), 3 qh )

(18)

vh ) + (∂x ph (t),3 vh ) + (dπhk−1 mh (t),3 vh ) = (g(t),3 vh ) (b∂t mh (t),3

(19)

3h = Pk−1 (Th ) and 3 3h = Pk−1 (Th ), and for all 0 ≤ t ≤ T . vh ∈ V for all 3 qh ∈ Q The well-posedness of this problem follows from the results of the next section.

4 Discrete Stability Estimates We now derive some discrete stability estimates that yield well-posedness of the semidiscrete method and that allow us to establish error estimates of optimal order. Lemma 1 Let ph , mh denote a solution of Problem 1. Then aπhk−1 ph (t )2 + bπhk−1 mh (t )2 # ≤ C(T )

aπhk−1 ph (0)2

+ bπhk−1 mh (0)2

 + 0

t

1 1 k−1 π f (s)2 + πhk−1 g(s)2 ds a h b

with constant C(T ) ≤ CT and C independent of T and the solution.

$

An Inexact Petrov-Galerkin Approximation for Gas Transport in Pipeline Networks

433

Proof Let us first note that (πhk−1 qh , πhk−1 q) = (qh , πhk−1 q) for all q ∈ H 1 (xL , xR ). By testing (18)–(19) with qh = πhk−1 ph (t) and vh = πhk−1 mh (t), we then get d dt



a k−1 b π ph (t)2 + πhk−1 mh (t)2 2 h 2

= (a∂t ph (t), πhk−1 ph (t)) + (b∂t mh (t), πhk−1 mh (t)) = −(∂x mh (t), πhk−1 ph (t)) − (∂x ph (t), πhk−1 mh (t)) − (dπhk−1 mh (t), πhk−1 mh (t)) + (πhk−1 f (t), πhk−1 ph (t)) + (πhk−1 g(t), πhk−1 mh (t)). By identity (17), integration-by-parts, and the boundary conditions (8), one can verify that (∂x mh (t), πhk−1 ph (t)) + (∂x ph (t), πhk−1 mh (t)) = 0. Via CauchySchwarz and Young inequalities, and using positivity of d, we then obtain the estimate d dt



a k−1 b ph (t)2 + πhk−1 mh (t)2 π 2 h 2

= −(dπhk−1 mh (t), πhk−1 mh (t)) + (πhk−1 f (t), πhk−1 ph (t)) + (πhk−1 g(t), πhk−1 mh (t)) ≤

α 1 1 k−1 1 (aπhk−1 ph (t)2 + bπhk−1 mh (t)2 ) + ( πh f (t)2 + πhk−1 g(t)2 ). 2 2α a b

The Gronwall lemma and the choice α = 1/T finally yields the assertion.

& %

Note that the above estimate does not yet give full control over the solution. A repeated application, however, allows us to prove the following stability estimate. Lemma 2 Let ph , mh denote a solution of Problem 1. Then ph (t)2 + mh (t)2  ≤ C  (T ) πhk−1 ph (0)2 + πhk−1 mh (0)2 + hπhk−1 ∂t ph (0)2 + hπhk−1 ∂t mh (0)2 +

 t 0

πhk−1 f (s)2 + πhk−1 g(s)2 ds + hπhk−1 ∂t f (s)2 + hπhk−1 ∂t g(s)2



for all 0 ≤ t ≤ T with C  (T ) = C  T and C  independent of T and of the solution. Proof As a direct consequence of the Poincaré inequality, one has ph  ≤ πhk−1 ph  + h∂x ph 

and mh  ≤ πhk−1 mh  + h∂x mh .

434

H. Egger et al.

The first terms in these estimates are already covered by Lemma 1. From the two Eqs. (18)–(19) with 3 qh = ∂x mh (t) and 3 vh = ∂x ph (t), we further deduce that ∂x mh (t)2 ≤ (πhk−1 f (t) + aπhk−1 ∂t ph (t))∂x mh (t)

and

∂x ph (t)2 ≤ (πhk−1 g(t) + bπhk−1 ∂t mh (t) + dπhk−1 mh (t))∂x ph (t). Bounds for πhk−1 ∂t ph (t) and πhk−1 ∂t mh (t) can be obtained by formally differentiating (18)–(19) with respect to time and applying Lemma 1 for the resulting system. A combination of the above estimates then yields the assertion of the lemma. & % Remark 1 Problem 1 formally amounts to a finite dimensional system of differential algebraic equations. From the stability estimates of Lemma 2 and [6, Theorem 4.12], one can deduce that this system is solvable for any choice of admissible initial values. The semidiscretization is thus well-defined. Further note that the stability constants in Lemma 1 and 2 are independent of the polynomial degree k.

5 Error Estimates As usual, we decompose the error according to p−ph  ≤ p−Ihk p+Ihk p−ph  and m − mh  ≤ m − Ihk m + Ihk m − mh  into approximation and discrete error components. The first part can be handled by the following estimates [11]. To simplify notation, we assume that the mesh is quasi-uniform in the following. Lemma 3 Let w ∈ H s+1(Th ), 0 ≤ s ≤ k. Then w − Ihk w ≤ C

 s+1 h k

|w|s+1;h .

(20)

For any w ∈ L2 (xL , xR ) ∩ H s (Th ), 0 ≤ s ≤ k, one has w − πhk−1 w ≤ C

 s h k

|w|s;h .

(21)

, xR ) : w|T ∈ H s (T )} is the space of piecewise smooth Here H s (Th ) = {w ∈ L2 (xL< functions and |w|s;h := ( T ∂xs w2L2 (T ) )1/2 is the corresponding seminorm. Moreover, the constant C in the estimates is independent of h and k. Using Eqs. (9)–(10) and (18)–(19) characterizing the continuous and the discrete solutions, one can see that the discrete error components p Bh (t) := Ihk p(t) − ph (t) Bh (0) = 0 and m Bh (t) := Ihk m(t) − mh (t) satisfy Eqs. (18)–(19) with initial values p

An Inexact Petrov-Galerkin Approximation for Gas Transport in Pipeline Networks

435

and m Bh (0) = 0, and right hand sides given by fB(t) := a(Ihk ∂t p(t) − ∂t p(t))

and

B g (t) := b(Ihk ∂t m(t) − ∂t m(t)) + d(πhk−1 Ihk m(t) − m(t)). By the a-priori estimates of Lemma 2, one then obtains the following result. Lemma 4 Let d ∈ P0 (Th ) be piecewise constant. Then for all 0 ≤ t ≤ T one has Ihk p(t) − ph (t)2 + Ihk m(t) − mh (t)2  ≤ C  (T ) hIhk ∂t p(0) − ∂t p(0)2 + hIhk ∂t m(0) − ∂t m(0)2  t + Ihk m(s) − m(s)2 + Ihk ∂t p(s) − ∂t p(s)2 + Ihk ∂t m(s) − ∂t m(s)2 0

 + hIhk ∂t t p(s) − ∂t t p(s)2 + hIhk ∂t t m(s) − ∂t t m(s)2 ds ,

with a constant C  (T ) = C  T and C  independent of h, k, T , and of the solution. Proof We apply Lemma 2 for p Bh (t) = Ihk p(t)−ph (t) and m Bh (t) = Ihk m(t)−mh (t) and then estimate the terms on the right hand side of the result step by step. By definition of the initial values, we have p Bh (0) = m Bh (0) = 0. Moreover, πhk−1 ∂t ph (0) = πhk−1 f (0) − ∂x mh (0) = πhk−1 f (0) − ∂x Ihk m(0) = πhk−1 f (0) − πhk−1 ∂x m(0) = πhk−1 ∂t p(0), where we used the definition of the initial value mh (0) in the second and (17) in the third step. Thus πhk−1 ∂t p Bh (0) ≤ Ihk ∂t p(0) − ∂t p(0), and in a similar k−1 manner, one can show πh ∂t m Bh (0) ≤ Ihk ∂t m(0) − ∂t m(0). This explains the first two terms in the estimate in the lemma. The terms under the integral are derived by estimating πhk−1 fB(t), πhk−1B g (t) and the derivatives πhk−1 ∂t fB(t), g (t) via the triangle inequality, and noting that πhk−1 ∂t B πhk−1 (dπhk−1 Ihk m(t) − dm(t)) = dπhk−1 (Ihk m(t) − m(t)), where we used that d is piecewise constant.

& %

Remark 2 A similar result can be proven for piecewise smooth d ∈ W 1,∞ (Th ) and additional terms of the form d −πh0 dπhk−1 p(t)−p(t) arise. For d ∈ W 1,∞ (Th ), the product of the two terms again has optimal approximation order. By combination of the above estimates, we finally obtain the following result.

436

H. Egger et al.

Theorem 1 Let (A1)–(A2) hold and d ∈ W 1,∞ (Th ). Furthermore, let (p, m) be a sufficiently smooth solution of (6)–(8). Then for all 0 ≤ t ≤ T , one has p(t) − ph (t) + m(t) − mh (t) ≤ C(u, p, T )

hk+1 . kk

For sufficiently smooth solutions, the proposed method thus converges at optimal order in h and at almost optimal order in the polynomial degree k.

6 Extension to Networks We now illustrate that our method and the convergence results of the previous section can be generalized easily to pipe networks. Let (V, E) denote a directed graph with vertices v ∈ V and edges e ∈ E; see Fig. 1 for illustration. For any edge e = (v1 , v2 ), we define ne (v1 ) = −1 and ne (v2 ) = 1. The matrix N with entries Nij = nej (vi ) then is the incidence matrix of the graph. For any vertex v ∈ V, we define E(v) = {e : e = (v, ·) or e = (·, v)}, and we set V0 = {v ∈ V : |E(v)| > 1} and V∂ = {v ∈ V : |E(v)| = 1} which gives a decomposition V = V0 ∪ V∂ into interior and boundary vertices. To every edge e, we associate a positive length e , and we identify e with [0, e ] in the sequel. This allows us to define spaces Lp (E) = {v : v|e ∈ Lp (e)} and H 1 (E) = {v ∈ Lp (E) : v|e ∈ H 1 (e)} of, respectively, integrable and piecewise smooth functions on the graph. The flow of gas in a pipe network is then described as follows: On every edge e representing a pipe, we require that a e ∂t pe + ∂x me = f e

(22)

b ∂t m + ∂x p + d m = g ,

(23)

e

e

e

e

e

e

where f e = f |e denotes the restriction of a function f ∈ Lp (E) to one edge. The equations for the individual pipes are coupled by algebraic conditions  e∈E(v)

me (v)ne (v) = 0 

pe (v) = pe (v)

v ∈ V0

(24)

v ∈ V0 , e, e ∈ E(v)

(25)

at the pipe junctions, and at the boundary vertices, we assume that pe (v) = 0

v ∈ V∂ .

(26)

An Inexact Petrov-Galerkin Approximation for Gas Transport in Pipeline Networks v3

e5

e2 v1

e1

v2

e3

v5

e4

Fig. 1 Directed graph (V, E) modeling the pipe network topology used for numerical tests

437

e7

v6

e6 v4

Inhomogeneous and other types of boundary conditions can again be incorporated with minor modifications. For the analysis of the problem, we now utilize the spaces H01 := {p ∈ H 1 (E) : (25) and (26) are valid}

(27)

H (div) := {m ∈ H 1 (E) : (24) is valid}

(28)

which are the natural generalization of those used for the analysis on a single pipe. Any solution (p, m) of (22)–(26) then again satisfies p(t) ∈ H01 , m(t) ∈ H (div), and q ) + (∂x m(t), 3 q ) = (f (t), 3 q) (a∂t p(t), 3

(29)

v ) + (∂x p(t),3 v ) + (dm(t),3 v ) = (g(t), 3 q) (b∂t m(t),3

(30)

for all 3 q ∈ L;2 (E), 3 v ∈ L2 (E), and all 0 < t < T . Here (v, w) = e e e (v , w )e = e v we dx denotes the scalar product on L2 (E).


0, the control problem (1) aims to determine the optimal state and control L. H. Christiansen () · J. B. Jørgensen Department of Applied Mathematics and Computer Science & Center for Energy Resources Engineering, Technical University of Denmark (DTU), Kgs. Lyngby, Denmark e-mail: [email protected]; [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_35

441

442

L. H. Christiansen and J. B. Jørgensen

variables, (y ∗ , u∗ ), that minimize the objective (1a). Here the optimal solution must belong to the set of feasible pairs, (y, u), that satisfy the PDE-constraints (1b) and the additional admissibility condition, u ∈ Uad . To be concrete, this paper focuses on the case of bi-lateral point-wise control constraints Uad := {u ∈ L2 (Ω) : ua ≤ u(x) ≤ ub a.e. in Ωd }.

(3)

Point-wise bounds of the type (3) appear in a number of practical applications, where the control must satisfy, e.g., operational limitations that are not naturally captured by the underlying PDE (1b). In the limiting case, where ua := −∞ and ub := ∞, the admissible set becomes Uad = L2 (Ωd ). This corresponds to the case where the PDE (1b) constitutes the only constraint.

1.1 Main Contributions and Outline This paper contributes to a recent series of efforts by the authors that seek to construct fast, iterative solvers for a range of PDE-constrained optimization problems by exploiting the properties of customized spectral bases [4–6]. This series of work aims to introduce a high-order alternative to the widely-used constellation of low-order finite-element methods and Schur-complement preconditioners that currently predominates the literature on PDE control [12–14]. Previous efforts have mainly considered distributed control of elliptic and parabolic non-linear diffusionreaction systems. The main focus has been on problems in rectangular domains, where PDEs constitute the only constraints. As a natural extension, this paper investigates how to modify the existing methods to account for (1) bound constraints of the type (3) and (2) different geometries. For the sake of brevity, the paper restricts attention to annular domains (2). However, with slight modifications, the approach generalizes to cylindrical geometries of the type ΩC := {(x, y, z) ∈ R3 | a ≤ x 2 + y 2 ≤ b, z ∈ (0, h)}, 0 < a < b.

(4)

As the main contribution, this work proposes a collection of Poisson-like preconditioners that are customized for efficient solution of the control problems (1) by a semi-smooth Newton (SSN) strategy [9]. Similar to a traditional Newton method, the SSN scheme solves (1) iteratively by finding a locally optimal solution to the non-linear Karuhn-Kush-Tucker (KKT) optimality conditions by solving a sequence of linearized, variable-coefficient subproblems. Direct solution of the subproblems is often time consuming and requires considerable memory-allocation. To this end, the new preconditioners are designed to promote efficient solution of the SSN subproblems by appropriate Krylov subspace (KSP) methods. Following seminal ideas of Shen [16], the preconditioners rely on fast direct solvers for constantcoefficient problems that exploit (1) the structure of boundary-adapted spectral bases and (2) the separable nature of annular domains. As the main feature, inversion

New Preconditioners for Semi-linear PDE-Constrained Optimization

443

of the preconditioners decouples to form to a sequence of independent 2×2 systems. This implies that the preconditioners can be applied matrix-free and scale linearly with the problem size. In addition, the independence of the 2 × 2 systems makes the preconditioners amenable to parallelization. To establish proof-of-concept, a numerical case study solves (1), where G(·) is given by a cubic non-linearity. The results demonstrate computational efficiency and show that the preconditioners respond well to different problem sizes, boundary conditions, point-wise bound constraints and various choices of the regularization parameter, ρ > 0. To establish the necessary background, Sect. 2 outlines how to solve the optimal control problem (1) using the SSN scheme. Further, to motivate the contributions of this paper, the section discusses some of the computational challenges that arise from discretization of the associated linearized subproblems. These challenges naturally leads to the construction of the new Poisson-like preconditioners in Sect. 3. Section 4 presents numerical results, while Sect. 5 draws overall conclusions and addresses future work.

2 Motivation: A Semi-smooth Newton Method This paper solves the control problem (1) by a semi-smooth Newton strategy [9]. The SSN scheme seeks to generate a locally optimal solution, (y, u), by solving the first-order necessary optimality system −Δy + G(y) − H (p) = 0

in Ω,

(5a)

−Δp + Gy (y)p + ϕy (y) = 0

in Ω.

(5b)

Here the boundary conditions of the original problem (1) are preserved, Gy denotes the Fréchet derivative of G with respect to the state variable, y, and the optimal control satisfies u = H (p) = max(ua , min(ρ −1 p(x), ub )). In the special case Uad := L2 (Ω), it can be shown that u = H (p) = ρ −1 p [18]. In the concrete case of annular domains (2), the system (5) can be recast to polar coordinates. To this end, define the functions Y (t, θ ) := y(r(t) cos(θ ), r(t) sin(θ )), P (t, θ ) := p(r(t) cos(θ ), r(t) sin(θ )), (6) where r(t) :=

b−a 2 (t + c),

t ∈ [−1, 1], c =

b+a b−a .

−Δt Y + κG(Y ) − κH (P ) = 0

The optimality system then reads

in

−Δt P + κGY (Y )P + κϕY (Y ) = 0 in

QR

(7a)

QR ,

(7b)

444

L. H. Christiansen and J. B. Jørgensen

  2 1 where Δt Y := ((t + c)Yt )t + (t +c) Yθθ , κ = (t +c)(b−a) and QR := [−1, 1] × 4 [0, 2π). To solve the KKT conditions (7), the SSN scheme considers the system as an operator equation F (y, p) = 0 and solves it by generating a recursive sequence of iterates, xi := (Yi , Pi ), 1 ≤ i ≤ k, where the next iterate, xk+1 := (Y, P ), is found by solution of the linearized optimality conditions: −Δt Y + C0 (xk )Y − C1 (xk )P = f (xk )

in Ω,

(8a)

−Δt P + C0 (xk )P + C2 (xk )Y = g(xk )

in Ω.

(8b)

Here C0 (xk ) := κGY (Yk ), C1 (xk ) := κHP (Pk ), C2 (xk ) := κ(GY Y (Yk )Pk + ϕY Y (Yk )) and f (xk ) := κ(GY (Yk )Yk − G(Yk ) − (HP (Pk )Pk − H (Pk ))),

(9a)

g(xk ) := κ(GY Y (Yk )Pk Yk + ϕY Y (Yk )Yk − ϕY (Yk )),

(9b)

where Hp denotes the generalized Newton derivative of H with respect to the adjoint variable, P , i.e., ⎧ 1 ⎨1 if ua ≤ ρ1 P ≤ ub , HP (P ) = ρ ⎩0 otherwise.

(10)

2.1 Numerical Challenges: Discretization of the SSN Subproblems As a numerical challenge, the SSN scheme relies on successive solution of coupled PDEs in the form (8). Upon discretization, this leads to repeated solution of large saddle-point problems. To illustrate the associated difficulties, consider a spectral-Galerkin discretization of the linear subproblems (8). To this end, define the boundary-adapted approximation spaces VN := {v ∈ PN : av(±1)+bv  (±1) = 0},

FM := span{eik(·) , M/2 ≤ k ≤ M/2−1}.

(11) Let K := N · M and define SK := VN × FM . The discrete Galerkin approximation of (8) then seeks to find Y, P ∈ SK such that (t + c)Yt , vt  + (t + c)−1 Yθ , vθ  + C0 Y − C1 P , v = f, v

∀v ∈ SK , (12a)

(t + c)Pt , vt  + (t + c)−1 Pθ , vθ  + C0 P + C2 Y, v = g, v

∀v ∈ SK , (12b)

New Preconditioners for Semi-linear PDE-Constrained Optimization



2π

where v, w :=



1 −1

0

445

vw dtdθ. To represent the approximate solutions, YN,M

and PN,M , consider the truncated series expansions YN,M (t, θ ) :=

M/2−1 −2  N

M/2−1 −2  N

B yl(k)m ψm (t )eikθ , PN,M (t, θ ) :=

k=−M/2 m=0

p Bl(k)m ψm (t )eikθ ,

k=−M/2 m=0

(13) where l(k) := k + the basis {ψk }N−2 k=0 :

M 2 .

Now, define the (N − 1) × (N − 1) matrices associated with

aij = (c + t)ψj , ψi ,

A = (aij )i,j =0..N−2 ,

(14)

bij = (c + t)−1 ψj , ψi ,

B = (bij )i,j =0..N−2 .

(15)

Note that appropriate choices of the basis functions {ψk }N−2 k=0 ∈ Vn will be constructed in Sect. 3. Further, let Γ and Ξ denote the M × M diagonal matrices defined by γmn = ein(·) , eim(·)  = 2πδmn , ξmn = mnein(·) , eim(·)  = 2πnmδmn ,

(16)

where δmn denotes the Kronecker delta. Finally, consider the (MN × 1) vectors B y := (B y0 , . . . , B yM−1 ), B yk = {B yj k }N−2 j =0 ,

(17)

p B := (B p0 , . . . , p BM−1 ), p Bk = {B pj k }N−2 j =0 ,

(18)

B := (B gM−1 ), B gk = {g, ψj eik(·)}N−2 G g0 , . . . , B j =0 ,

(19)

B := (fB0 , . . . , fBM−1 ), fBk = {f, ψj eik(·) }N−2 . F j =0

(20)

The discretized linear subproblem (8) can then be written in matrix form :9 : 9 : B B y MC2 B + MC0 G = B , p B B + MC0 −MC1 F F GH I F GH I F GH I 9

A

x

(21)

b

where B = Γ ⊗ A + Ξ ⊗ B. Here the matrices MC ,  = 1, 2, 3 are defined by the elements (mc )ij = Ci ψk eim(·), ψl ein(·) ,

(22)

446

L. H. Christiansen and J. B. Jørgensen

where i, j satisfy that i = n(N − 1) + (l + 1), j = m(N − 1) + (k + 1),

(23a)

0 ≤ k, l ≤ N − 2,

(23b)

0 ≤ n, m ≤ M − 1.

3 New Poisson-Like Preconditioners As a significant challenge to the numerical solution of (7), the SSN scheme relies on repeated solution of saddle-point problems (21) of dimension 2(N − 1)M × 2(N − 1)M. Consequently, direct solution strategies often become computational intractable. As a cost efficient alternative, the following introduces new preconditioners that seek to accelerate the inner SSN subproblems (8) by using appropriate Krylov subspace methods to solve the associated preconditioned linear systems Pk−1 Ak xk = Pk−1 bk .

(24)

Concretely, this paper proposes approximative constraint preconditioners of the type : BC0 BC2 B B+M M Pk = B B BC1 . B + MC0 −M 9

(25)

Following ideas of traditional Poisson preconditioners, the new preconditioners are constructed by approximating each block of the SSN subproblem (21) by the Bc ,  = 0, 1, 2, that come from a spectral Galerkin discretization matrices, B B and M of the corresponding constant-coefficient problem that determines Y, P ∈ SK such that CA Yt , vt  + CB Yθ , vθ  + C 0 Y − C 1 P , v = f, v

∀v ∈ SK ,

(26a)

CA Pt , vt  + CB Pθ , vθ  + C 0 P + C 2 Y, v = g, v

∀v ∈ SK ,

(26b)

where CA = c, CB =

c and C i = 2 c −1

 1 2

max Ci (xk ) + min Ci (xk ) , i = Ω

Ω

0, 1, 2. To be efficient, the new preconditioners crucially rely on carefully chosen basis functions {ψk }N−2 k=0 for the discrete approximation space, VN (11). To this end, this paper uses Fourier-like (FL) bases that were originally introduced by Shen and Wang in the context of traditional initial-boundary-value problems [17]. As a key property to construction of the preconditioners, the FL bases lead to diagonal massand stiffness matrices, i.e., Mij = (ψj , ψi )ij = λj δj,i , Sij = (∂t ψj , ∂t ψi )ij = δj,i .

(27)

New Preconditioners for Semi-linear PDE-Constrained Optimization

447

The FL bases can be constructed as part of an offline preprocessing stage in two steps: 1. Let {Lk (·)}N k=0 be the Legendre polynomials. Then there exists a unique set of coefficients {ak , bk }N−2 k=0 such that   φk := ck Lk + ak Lk+1 + bk Lk+2 ∈ Vk+2 ,

, ck := ( −bk (4k + 6))−1 .

Furthermore, the mass matrix, MA = (φj , φi )ij , is penta-diagonal and symmetric positive definite, whereas the stiffness matrix, SA = (∂x φj , ∂x φi )ij , becomes diagonal [15]. In the concrete cases of Dirichlet and Neumann boundary conditions, the coefficients, {ak , bk }N−2 k=0 are given by respectively ak = 0, bk = −1 and ak = 0, b0 = 1/2, bk = −k(k + 1)/((k + 2)(k + 3)). (28) 2. The second step computes the diagonalization Λ = QT MA Q, where Q = (qij ) denotes the matrix of eigenvectors and {λi }N−2 i=1 are the associated eigenvalues. Using the matrix Q, the FL basis can be constructed by the linear combinations: ψk (x) =

N−2 

qj k φj (x), 0 ≤ k ≤ N − 2.

(29)

j =0

3.1 Efficient Inversion of the Preconditioners As the main feature of the preconditioners, Pk , the following describes an efficient inversion procedure that exploits the orthogonal structures of the FL bases (27). To this end, consider the following preconditioning problem that is solved during each iteration of the KSP method: 9 :9 : 9 : BC2 B Bk BC0 M G B yk B+M = Bk . (30) B BC1 BC0 −M p Bk F B+M F GH I F GH I F GH I Pk

zk

Ak xk

448

L. H. Christiansen and J. B. Jørgensen

Note that (30) corresponds to the discrete first-order necessary optimality conditions associated with the constant-coefficient optimal control problem (26). Hence, by definition (22), it follows that B BC = C  Γ ⊗ M, B = CA Γ ⊗ S + CB Ξ ⊗ M, M

(31)

where Sij = (∂t ψj , ∂t ψi )ij and Mij = (ψj , ψi )ij . Further, by the orthogonal properties of the Fourier bases (16), the matrices, Γ and Ξ , are diagonal. Therefore, using the notation, k N−2 k N−2 Bk k N−2 Bklm }N−2 , F Blk = {F Blm B ylk = {B ylm }m=0 , p Blk = {B plm }m=0 , Gl = {G }m=0 , m=0

it follows that the preconditioning problem (30) can be written as M independent linear systems 9

2πC2 M Σl Σl −2πC1 M

:9

B ylk p Blk

:

9 =

Bk G l Bk F l

: , 0 ≤ l ≤ M − 1,

(32)

where Σl := CA S + (CB k(l)2 + 2πC0 )M. In addition, the properties of the FL basis, {ψk }N−2 k=0 , implies that S and M become diagonal (29). Hence, the system (32) reduces to M(N − 1) independent 2 × 2 linear systems in the form 9

:9 2πC 2 λm σnm σnm −2πC1 λm

k B ylm k p Blm

:

9 =

: Bknm G , 0 ≤ l ≤ M − 1, 0 ≤ m ≤ N − 2, k Bnm F (33)

where σlm := CA + (CB k(l)2 + 2πC0 )λm . By (33), it follows that the original preconditioning problem (30) decouples into (N − 1)M independent 2 × 2 subsystems. As a consequence, the Poisson-like preconditioners (25) scale linearly with the problem size and can be applied matrix-free.

4 Numerical Results To investigate the potential of the Poisson-like preconditioners, the following case study solves the control problem (1), where the reaction term is given by the cubic non-linearity G(y) := y 3 . The corresponding problem serves as a recurring example

New Preconditioners for Semi-linear PDE-Constrained Optimization

449

in the control literature [18]. In this case study, the goal is to track the desired state of the type ⎧ ⎨Z, (r, θ ) ∈ [α, β] × [0, π/2] ∪ [π, π/3] zd (r, θ ) = ⎩0, otherwise

,

(34)

where a ≤ α < β ≤ b. The following example uses the parameters, Z = 4, a = 30, α = 40 and β = b = 60. The main purpose of the study is to investigate efficiency and robustness of the preconditioners (25). To this end, the study solves (1) for different choices of (1) problem size, (2) boundary conditions, (3) regularization parameter, and (4) point-wise bound constraints of the type (3).1 As a benchmark reference, the results are compared to MATLABs state-of-the-art direct solver. All computations are carried out in [11] on a 2.9 GHz Intel processor. The SSN scheme is said to have converged when the 2-norm difference between successive iterates is below η = 10−4 . The KSP iterations are performed using the MATLAB function GMRES with a tolerance of # = 10−9 . The direct solver relies on MATLABs backslash command. Table 1 lists the results, where KSP iter denotes the average number of KSP iterations required for each SSN step. Note also that DOF denotes the number of degrees of freedom for each individual SSN subproblem. Hence, the total degrees of freedom, DOFT , is therefore given by #SSN steps × DOF. The results reflect some overall tendencies that generalize to other choices of the parameters, Z, a, α, β and b. Firstly, the preconditioners provide significant reductions in CPU-time compared to the direct strategy. In particular, the results show that the non-linear control problem with up to DOFT = 875,000 unknowns can be solved in less than a minute using modest hardware. Secondly, the preconditioners prove robust with respect to the problem size and the choice of boundary conditions. Thirdly, as a drawback, the number of SSN steps and KSP iterations increase as the point-wise bounds become more strict. The authors suspect that these increases in SSN steps and KSP iterations are caused by the combination of a decrease in regularity of the solution and an increase in nonlinearity of the KKT system (Fig. 1).

1 By the choices of parameters, the study strives to provide a representative example of the general tendencies of performance and robustness that can be expected from the preconditioners. To allow for more diverse and elaborate experiments, the MATLAB source code of this study has been made publicly available from https://github.com/LHCH-DK/PDE_Control_Annular.git.

24 26 27

1.77 [86.51] 5.42 [–] 51.81 [–]

6 6 6

7 7 7

28 29 32

28 30 33

1.57 [83.91] 4.25 [–] 34.23 [–]

7 7 7

1.71 [82.27] 5.53 [–] 51.9 [–]

6 6 6

1.42 [76.13] 4.26 [–] 36.35 [–]

25 26 28

ua = −10, ub = 10, ρ = 10−3 Time (s) SSN steps KSP iter.

ua = −35, ub = 35, ρ = 10−4 Time (s) SSN steps. KSP iter.

A horizontal lines indicates that the computations were manually terminated after 300 s, without reaching convergence

ua = −∞, ub = ∞, ρ = 10−5 N,M DOF Time (s) SSN steps KSP iter. Optimal tracking problem: Dirichlet boundary conditions 50 5000 0.54 [59.95] 4 9 100 20,000 1.23 [–] 4 9 250 125,000 7.21 [–] 4 9 Optimal tracking problem: Neumann boundary conditions 50 5000 0.65 [53.26] 4 10 100 20,000 1.25 [–] 4 10 250 125,000 7.43 [–] 4 8

Table 1 For comparison, [·] denotes the CPU time (s) required by MATLABs direct solver

450 L. H. Christiansen and J. B. Jørgensen

New Preconditioners for Semi-linear PDE-Constrained Optimization

451

Fig. 1 The computed states for (1) Dirichlet boundary conditions, (2) Neumann boundary conditions and (3) the desired state for ua = −35, ub = 35, ρ = 10−4 . Note that both solutions manage to approximate the desired state well, despite of the bound constraints

5 Conclusions and Outlook This paper has proposed new Poisson-like preconditioners for semi-linear PDEconstrained optimization problems with non-linear reaction kinetics and point-wise bound constraints. The preconditioners specifically target problems in annular domains. Inspired by [16], the new preconditioners exploit the orthogonal properties of customized, boundary-adapted spectral bases. This leads to matrix-free preconditioners that scale linearly with the problem size. Numerical results have demonstrated that the preconditioners lead to fast solution of large-scale optimization problems with significant computational benefits compared to MATLABs state-of-the-art direct methods. Furthermore, the preconditioners have proven to be robust with respect to the problem size for both homogeneous Dirichlet and Neumann boundary conditions. As a challenge, numerical experiments indicated that the non-linearity of the problem increases as the point-wise bound constraints become more strict. In turn, this leads to an increase in the number of SSN steps and KSP iterations that are required to reach convergence. A future study seeks to improve this situation by providing the SSN scheme with an educated starting guess that uses a coarse-grid solution to a similar control problem with less restrictive constraints.

References 1. Biegler, L.: Efficient solution of dynamic optimization and NMPC problems. Prog. Syst. C 26, 219–243 (2000) 2. Biegler, L.T.: Real-Time PDE-Constrained Optimization, vol. 3. Society for Industrial and Applied Mathematics, Philadelphia (2007) 3. Borzi, A.: Multigrid methods for parabolic distributed optimal control problems. J. Comput. Appl. Math. 157(2), 365–382 (2003)

452

L. H. Christiansen and J. B. Jørgensen

4. Christiansen, L.H., Jørgensen, J.B.: A fast and memory-efficient spectral Galerkin scheme for distributed elliptic optimal control problems (2017). Preprint. arXiv:1712.08225 5. Christiansen, L.H., Jørgensen, J.B.: A fast PDE-constrained optimization solver for nonlinear diffusion-reaction processes. In: Proceedings of 2018 IEEE Conference on Decision and Control, pp. 2635–2640. IEEE (2018) 6. Christiansen, L.H., Jørgensen, J.B.: A new Lagrange-Newton-Krylov solver for PDEconstrained nonlinear model predictive control. IFAC-PapersOnLine 51(20), 325–330 (2018) 7. Diehl, M., Bock, H., Schloder, J.: Newton-type methods for the approximate solution of nonlinear programming problems in real-time. Appl. Optim. 82, 177–200 (2003) 8. Herzog, R., Kunisch, K.: Algorithms for PDE-constrained optimization. Gamm Mitteilungen 33(2), 163–176 (2010) 9. Ito, K., Kunisch, K.: Lagrange Multiplier Approach to Variational Problems and Applications. Society for Industrial and Applied Mathematics, Philadelphia (2008) 10. Leugering, G., Benner, P., Engell, S., Griewank, A., Harbrecht, H., Hinze, M., Rannacher, R., Ulbrich, S.: Trends in PDE Constrained Optimization. International Series of Numerical Mathematics. Springer International Publishing, Basel (2014) 11. MATLAB.: version 8.6.0 (R2015b). The MathWorks Inc., Natick, Massachusetts (2015) 12. Pearson, J.W., Stoll, M.: Fast iterative solution of reaction-diffusion control problems arising from chemical processes. SIAM J. Sci. Comput. 35(5), B987–B1009 (2013) 13. Rees, T., Wathen, A.J.: Preconditioning iterative methods for the optimal control of the Stokes equations. SIAM J. Sci. Comput. 33(5), 2903–2926 (2011) 14. Rees, T., Dollar, H.S., Wathen, A.J.: Optimal solvers for PDE-constrained optimization. SIAM J. Sci. Comput. 32(1), 271–298 (2010) 15. Shen, J.: Efficient spectral-Galerkin method. I. Direct solvers of second- and fourth-order equations using Legendre polynomials. SIAM J. Sci. Comput. 15(6), 1489–1505 (1994) 16. Shen, J.: Efficient spectral-Galerkin methods III: polar and cylindrical geometries. SIAM J. Sci. Comput. 18(6), 1583–1604 (1997) 17. Shen, J., Wang, L.L.: Fourierization of the Legendre-Galerkin method and a new space-time spectral method. Appl. Numer. Math. 57(5–7), 710–720 (2007) 18. Tröltzsch, F.: Optimal Control of Partial Differential Equations: Theory, Methods, and Applications. Graduate Studies in Mathematics. American Mathematical Society, Providence (2010)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

DIRK Schemes with High Weak Stage Order David I. Ketcheson, Benjamin Seibold, David Shirokoff, and Dong Zhou

1 Introduction Runge-Kutta (RK) methods achieve high-order accuracy in time by means of combining approximations to the solution at multiple stages. An s-stage RK scheme can be represented via the Butcher tableau c1 .. .

a11 · · · .. .

a1s .. .

c A = . bT cs as1 · · · ass b1 · · · bs

D. I. Ketcheson Applied Mathematics and Computational Science, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia e-mail: [email protected] B. Seibold () Department of Mathematics, Temple University, Philadelphia, PA, USA e-mail: [email protected] D. Shirokoff Department of Mathematical Sciences, New Jersey Institute of Technology, Newark, NJ, USA e-mail: [email protected] D. Zhou Department of Mathematics, California State University Los Angeles, Los Angeles, CA, USA e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_36

453

454

D. I. Ketcheson et al.

Throughout the whole paper we assume that c = Ae, where e is the vector of all ones. The scheme’s stability function [12] R(ζ ) = 1 + ζ b T (I − ζ A)−1 e measures the growth un+1 /un per step Δt, when applying the scheme to the linear model equation u (t) = λu, with ζ = λΔt. A particular interest lies in the accuracy of the RK scheme for stiff problems, i.e., problems in which a larger time step is chosen than the fastest time scale of the problem’s dynamics. A standard stiff model problem [8] is the scalar linear ordinary differential equation (ODE) u = λ(u − φ(t)) + φ  (t) ,

(1)

with i.c. u(0) = φ(0) and Re λ ≤ 0. The true solution y(t) = φ(t) evolves on an O(1) time scale. Hence, λ-values with large negative real part result in stiffness. Considering a family of test problems (parametrized by λ), one can now establish the scheme’s convergence via two different limits: (a) the non-stiff limit Δt → 0 and ζ → 0; and (b) the stiff limit Δt → 0 and ζ → −∞. A characteristic property of most RK schemes is that, while the non-stiff limit recovers the scheme’s order (as given by the order conditions [2, 5]), the error decays at a reduced order in the stiff limit. This phenomenon is called “order reduction” (OR) [1, 3, 7, 10, 11] and it manifests in various ways for more complex problems, including numerical boundary layers [6]. The OR phenomenon can be seen by studying the RK scheme applied to (1). The approximation error at time tn+1 reads [12, Chapter IV.15] # n+1 = R(ζ ) # n + ζ b T (I − ζ A)−1 δ sn+1 + δ n+1 ,

(2)

where R(ζ ) is the growth factor, and δ sn+1 =

 j ≥2

Δt j (j −1)!

τ (j ) φ (j ) (tn ) ,

δ n+1 =

 j ≥1

Δt j (j −1)!

  b T c j −1 − j1 φ (j ) (tn )

are the truncation errors incurred at the intermediate stages and at the end of the step, respectively. Here, φ (j ) denotes the j -th derivative of the solution, and the vectors τ (j ) = Ac j −1 − j1 c j ,

j = 1, 2, . . .

we call the stage order residuals or stage order vectors. The condition τ (η) = 0 for 0 ≤ η ≤ j appears often in the literature and is also referred to as the simplifying assumption C(η) [12]. In (2), the step error δ n+1 is of the formal order (in Δt) of the scheme (due to the order conditions). Moreover, the growth factor carries over (more or less, see [4]) the accuracy from one to the next step. Hence, the critical expression for OR is the term involving the stage error δ sn+1 . Specifically, the asymptotic behavior of the expression g (j ) = ζ b T (I − ζ A)−1 τ (j )

(3)

DIRK Schemes with High Weak Stage Order

455

matters. In the non-stiff limit (ζ 1), a Neumann expansion yields ζ(I − ζ A)−1 = 2 3 2 ζ I + ζ A + ζ A + . . . , leading to expressions b T A τ (j ) with  > 0. And in fact the order conditions guarantee that bT A τ (j ) = 0 for 0 ≤  + j ≤ p − 1 to ensure the formal order of the scheme. Conversely, in the stiff limit we can treat ζ −1 as the small parameter and expand ζ (I −ζ A)−1 = −A−1 (I −ζ −1 A−1 )−1 = −A−1 −ζ −1 A−2 −ζ −2 A−3 −. . . , leading to expressions b T A τ (j ) with  < 0. The order conditions do not imply that these quantities vanish, and in general one may observe a reduced rate of convergence. A key question is therefore whether additional conditions can be imposed on the RK scheme that recover the scheme’s order in the stiff regime. A well-known answer to the question is: Definition 1 Let pˆ denote the order of the quadrature rule of an RK scheme. Let qˆ denote the largest integer such that τ (j ) = 0 for 1 ≤ j ≤ q. ˆ The stage order of a RK scheme is q = min(p, ˆ q). ˆ Having stage order q implies that the error decays at an order of (at least) q in the stiff regime (see also [12]). This work focuses particularly on diagonally-implicit Runge-Kutta (DIRK) schemes, for which A is lower diagonal. A known drawback of DIRK schemes is that they cannot have high stage order: Theorem 1 The stage order of an irreducible DIRK scheme is at most 2. The stage order of a DIRK scheme with non-singular A is at most 1. (2)

Proof Since c = Ae, we have τ 1 = a11 c1 − 12 (c1 )2 = 12 (a11 )2 . Thus if A is non-singular, one has τ (2) = 0, so q ≤ 1. Consider now the case that a11 = c1 = 0, and suppose that the method has stage order 3. The conditions τ 2(2) = τ 2(3) = 0 then imply a21 = a22 = c2 = 0, which would render the scheme reducible. Hence, q ≤ 2. & % Hence, while DIRK schemes possess an implementation-friendly structure (each stage is a backward-Euler-type solve), their potential to avoid OR by means of high stage order is limited. We therefore move to a weaker condition that can avoid OR in some situations for higher order in the context of DIRK schemes.

2 Weak Stage Order To avoid order reduction, the expressions g (j ) in (3) need to vanish in the stiff limit. In line with [9], we define the following criteria: Definition 2 (Weak Stage Order) A RK scheme has weak stage order (WSO) q˜ if there is an A-invariant subspace that is orthogonal to b and that contains the stage ˜ order vectors τ (j ) for 1 ≤ j ≤ q.

456

D. I. Ketcheson et al.

Theorem 2 (WSO Is the Most General Condition that Ensures g (j ) = 0 for All ζ > 0) Let coefficients A, b be given. Then g (j ) = 0 for all ζ > 0 and 1 ≤ j ≤ q˜ if and only if the corresponding RK scheme has weak stage order q. ˜ Proof Let C(G) denote the column space of ! " ˜ . G := τ (1) , Aτ (1), A2 τ (1) , . . . , As−1 τ (1) , τ (2) , Aτ (2) , . . . , As−1 τ (q) From the Cayley-Hamilton theorem it follows that WSO q˜ is equivalent to bT A τ (j ) = 0,

0 ≤  ≤ s − 1, 1 ≤ j ≤ q˜ .

(4)

⇒ Because C(G) is A-invariant, C(G) is invariant under multiplication by (1 − ζ A)−1 , i.e. if v ∈ C(G) then for any ζ > 0, the product (1 − ζ A)−1 v ∈ C(G). Since b is orthogonal to C(G), we have g (j ) = 0 for all 1 ≤ j ≤ q. ˜ ⇐ If g (j ) = 0, then ζ −1 g (j ) = bT (1 − ζ A)−1 τ (j ) = 0 for all ζ > 0. Differentiating both sides of this equation -times, with respect to ζ , and taking the limit as ζ → 0+ , yields the conditions in Eq. (4). & % Definition 3 (Weak Stage Order Eigenvector Criterion) A RK scheme satisfies the WSO eigenvector criterion of order q˜e if for each 1 ≤ j ≤ q˜e , there exists μj such that Aτ (j ) = μj τ (j ) , and moreover, b T τ (j ) = 0. The WSO eigenvector criterion of order q˜e implies WSO (of at least) q˜e . For a given scheme, let p denote the classical order, q the stage order, and q˜ the weak stage order. Then we have q˜ ≥ q and p ≥ q. Note however that a method with WSO q˜ ≥ 1 need not even be consistent; order conditions must be imposed separately. The WSO eigenvector criterion may serve to avoid OR because it implies that g (j ) = ζ b T (1 − ζ μj )−1 τ (j ) =

ζ b T τ (j ) , 1 − ζ μj

i.e., it allows one to “push” the stage order residuals past the matrix (1 − ζ A)−1 , and then use b T τ (j ) = 0. Note that the condition b T τ (j ) = 0 that is required in Definition 3 is actually automatically satisfied (due to the order conditions) if p > q˜e (or p ≥ q˜e for stiffly accurate schemes). It must be stressed that the concept of WSO (both criteria) is based on the linear test equation (1), hence it is not clear to what extent WSO will remedy OR for nonlinear problems or problems with time-dependent coefficients. In Sect. 4 we numerically investigate some nonlinear test problems. Finally, we present a limitation theorem on the WSO eigenvector criterion. Theorem 3 DIRK schemes with invertible A have q˜e ≤ 3. Proof Because the τ (j ) only depend on A, the eigenvector relation in Definition 3 depends only on A, not on b. With A lower triangular, the first k components

DIRK Schemes with High Weak Stage Order

457

of τ (j ) depend only on the upper k rows of A; and the same is true for the eigenvector relation as well. Hence, for a scheme to have an A that allows for the WSO eigenvector criterion of order q˜e , all upper sub-matrices of A must admit the same, too. We can therefore study A row by row. The first component of τ (j ) j equals (1 − j1 )a11, which is nonzero for j > 1. Hence, the first row of the equation Aτ (j ) = μj τ (j ) is equivalent to μj = a11 . With that, we can move to the second row of the equation, which reads   j j −1 (1− j1 )a11 a21 + (a22 −a11) a11 a21 + (a21 +a22)j −1 a22 − j1 (a21 +a22)j = 0 . (5) To determine the set of solutions (a11 , a21 , a22 ) of (5), we first observe that (5) is homogeneous, i.e., if (a11 , a21 , a22 ) solves (5), then (μa11 , μa21 , μa22 ) solves (5) as well for any μ ∈ R. It therefore suffices to consider the solutions of (5) in the 2D-plane ( aa11 , a22 ). Figure 1 shows the resulting solution curves for j ∈ {2, 3, 4}. 21 a21 One class of solutions lies on the straight line of slope 1 passing through (1, 0). Those schemes are equal-time methods, i.e., RK schemes that have c = νe, where ν ∈ R is a constant. In fact, equal-time schemes satisfy the eigenvector relation for all j . However, they are not particularly useful RK methods, because—among other limitations—they are restricted to second order. This follows because the order 1 and 2 conditions require b T e = 1 and b T c = 12 . Thus ν = 12 , and b T c2 = ν 2 = 14 , which contradicts the order 3 condition b T c2 = 13 . Note that the equal-time scenario also covers the points at infinity in Fig. 1, i.e., the schemes with a21 = 0. 1

10

0.8

6

0.6

4

0.4

2

0.2

a22/a21

a22/a21

8

order 2 & order 3 order 2 order 3 order 4

0

0

-2

-0.2

-4

-0.4

-6

-0.6

-8

-0.8

-10 -10

-8

-6

-4

-2

0

a11/a21

2

4

6

8

10

order 2 & order 3 order 2 order 3 order 4

-1

-1

-0.8 -0.6 -0.4 -0.2

0

0.2

0.4

0.6

0.8

1

a11/a21

Fig. 1 Curves of WSO orders 2, 3, and 4 as functions of the re-scaled parameters aa11 and aa22 . Left 21 21 panel: scale 10; right panel: scale 1. All orders are satisfied along the line of slope 1 going through (1,0), corresponding to equal-time DIRK schemes. Moreover, there are two further points (other than the origin), where orders 2 and 3 are satisfied. Neither of these two points satisfies order 4

458

D. I. Ketcheson et al.

Non-equal-time schemes that satisfy (5) for j √ = 2√ and j = 3 are the following a22 two points in the ( aa11 , ) plane: P = (−4 + 3 2, 2 − 1) = (0.2426, 0.4142) 1 a21 √ 21 √ √ and P2 = (−( 2 + 1)( 2 + 2), −( 2 + 1)) = (−8.2426, −2.4142). None of these two points satisfies (5) for j = 4 (green curve in Fig. 1). Therefore q˜e ≤ 3. & % Among the two sets of solutions found in the proof, P1 implies that a11, a21 , and a22 all have the same sign, which is a desirable property. In contrast, P2 implies that a21 < 0. Both WSO 3 schemes presented below correspond to the P1 solution.

3 DIRK Schemes with High Weak Stage Order Imposing the classical order conditions [2, 5], together with the WSO eigenvector relation (Definition 3), we determine RK schemes by searching the parameter space of DIRK schemes (with all diagonal entries non-zero). A stiffly accurate structure (b T equals the last row of A) is imposed, as is A-stability (verified by evaluating the stability function R(ζ ) along the imaginary axis). Together this implies that the resulting scheme is L-stable; i.e., it ensures that unresolved stiff modes decay [5]. The number of stages is chosen so that the constraints admit solutions. The optimization itself is carried out using MATLAB’s optimization toolbox, using multiple local optimization algorithms included in the function fmincon. An effort was made to minimize the L2 norm of the local truncation error coefficients. However, in multiple cases the solver exhibited bad convergence properties; so while the schemes below yield reasonable truncation errors, it should not be expected that they are optimal. We find an order 3 scheme with WSO 2 (see also [9]), 0.01900072890 0.78870323114 0.41643499339 1

0.01900072890 0.40434605601 0.38435717512 0.06487908412 −0.16389640295 0.51545231222 0.02343549374 −0.41207877888 0.96661161281 0.42203167233 0.02343549374 −0.41207877888 0.96661161281 0.42203167233

an order 3 scheme with WSO 3, 0.13756543551 0.13756543551 0.80179011576 0.56695122794 0.23483888782 2.33179673002 −1.08354072813 2.96618223864 0.44915521951 1 0.59761291500 −0.43420997584 −0.05305815322 0.88965521406 0.59761291500 −0.43420997584 −0.05305815322 0.88965521406

DIRK Schemes with High Weak Stage Order

459

and an order 4 scheme with WSO 3, 0.079672377876931 0.079672377876931 0 0 0 0 0.464364648310935 0.328355391763968 0.136009256546967 0 1.348559241946724 −0.650772774016417 1.742859063495349 0.256472952467792 0 1.312664210308764 −0.714580550967259 1.793745752775934 −0.078254785672497 0.311753794172585 0.989469293495897 −1.120092779092918 1.983452339867353 3.117393885836001 −3.761930177913743 1 0.214823667785537 0.536367363903245 0.154488125726409 −0.217748592703941 1 0.214823667785537 0.536367363903245 0.154488125726409 −0.217748592703941

0 0 0 0 0.770646024799205 0.072226422925896 0.072226422925896

0 0 0 0 0 0.239843012362853 0.239843012362853

4 Numerical Results In this section we verify the order of accuracy of the schemes above and demonstrate that WSO remedies order reduction for linear problems. We confirm that WSO p is required for ODEs, and WSO p − 1 is required for PDE IBVPs. In addition, we study the effect of WSO for two nonlinear problems.

4.1 Linear ODE Test Problem We consider the linear ODE test problem (1) with the true solution φ(t) = sin(t + π π 4 4 ), the stiffness parameter λ = −10 , and the initial condition u(0) = sin( 4 ). The problem is solved using three 3rd order DIRK schemes (with WSO 1, 2, and 3) and two 4th order DIRK schemes (with WSO 1 and 3)1 up to the final time T = 10. The convergence results are shown in Fig. 2. In the stiff regime where |ζ | = |λ|Δt ' 1, first order convergence is observed for the WSO 1 schemes as expected, the WSO 2 scheme improves the convergence rate to 2, and the WSO 3 schemes exhibit 3rd order convergence. In addition to yielding better convergence orders in the stiff regime, the schemes with higher WSO also turn out to yield substantially smaller error constants in the non-stiff regime (Δt 1/|λ|). For comparison, we also display a DIRK scheme with explicit first stage (EDIRK), that is, a11 = 0, of stage order 2 (see Theorem 1). The left panel of Fig. 2 shows that the WSO 2 scheme exhibits the same convergence behavior as the stage order 2 EDIRK scheme and performs equally well in terms of accuracy.

4.2 Linear PDE Test Problem: Schrödinger Equation As a linear PDE test problem, we study the dispersive Schrödinger equation. The method of manufactured solutions is used, i.e., the forcing, the boundary conditions (b.c.) and initial conditions (i.c.) are selected to generate a desired true solution. The

1 We

do not construct an order 4 scheme with WSO 2, as we see no role for such a method.

460

D. I. Ketcheson et al. Linear ODE test problem,

= -10 4, DIRK3 WSO1, 2 & 3

10-4

10-6

10-6

10-8

10-8

Error

Error

10-4

10-10

10-12

10-14

10-14

10

10-6

10-4

10-2

dt

10-16 10-6

100

= -10 4, DIRK4 WSO1 & 3

10-10

10-12

-16

Linear ODE test problem,

10-4

10-2

dt

100

Fig. 2 Error convergence for linear ODE test problem (1). Left: 3rd order DIRK schemes with WSO 1 (blue circles), WSO 2 (red triangles), WSO 3 (black squares), and a 3rd order EDIRK scheme with stage order 2 (light red dots). Right: 4th order DIRK schemes with WSO 1 (blue circles) and WSO 3 (red triangles)

10 -6 10 -8

10 -10 -4 10

10 -3

dt

10 -2

10 -1

10

-2

10

-4

Error in Maximum norm

10

-4

Error in Maximum norm

Error in Maximum norm

10

Schroedinger Eqn. DIRK4, WSO3

Schroedinger Eqn. DIRK3, WSO3

Schroedinger Eqn. DIRK3, WSO1 -2

10 -6 10 -8

10 -10 -4 10

10-3

dt

10 -2

10 -1

10

-2

10 -4 10 -6 10 -8

10 -10 -4 10

10 -3

dt

10 -2

10 -1

Fig. 3 Error convergence for the Schrödinger equation using 3rd order DIRK schemes with WSO 1 (left) and WSO 3 (middle), and a 4th order DIRK with WSO 3 (right)

spatial approximation is carried out using 4th order centered differences on a fixed spatial grid of 10,000 cells. This renders spatial approximation errors negligible and thus isolates the temporal errors due to DIRK schemes. The errors are measured in the maximum norm in space. We consider ut =

iω uxx for (x, t) ∈ (0, 1) × (0, 1.2], k2

u = g on {0, 1} × (0, 1.2] ,

(6)

with the true solution u(x, t) = ei(kx−ωt ), ω = 2π and k = 5. Figure 3 shows the convergence orders of u, ux and uxx for 3rd order DIRK schemes with WSO 1 (left), WSO 3 (middle) and a 4th order DIRK scheme with WSO 3 (right). For IBVPs, spatial boundary layers are produced by RK methods, thus limiting the convergence order in u to q˜ + 1, with an additional half an order loss per derivative when q˜ < p [9]. As a result, the 4th order WSO 3 scheme recovers 4th order convergence in u and improves the convergence in ux and uxx . When q˜ = p, the full convergence order in u, ux and uxx is achieved, as seen in the middle panel in Fig. 3.

DIRK Schemes with High Weak Stage Order Viscous Burgers Eqn. DIRK3, WSO1

10 -2

10 -4 10 -6 10 -8

10 -3

dt

10 -2

10 -1

Error in Maximum norm

10 -2

Error in Maximum norm

Error in Maximum norm

Viscous Burgers Eqn. DIRK3, WSO3

Viscous Burgers Eqn. DIRK3, WSO2

10 -2

10 -10 10 -4

461

10 -4 10 -6 10 -8

10 -10 10 -4

10 -3

dt

10 -2

10 -1

10 -4 10 -6 10 -8

10 -10 10 -4

10 -3

dt

10 -2

10 -1

Fig. 4 Error convergence for the viscous Burgers’ equation using 3rd order DIRK schemes with WSO 1 (left), WSO 2 (middle) and WSO 3 (right)

4.3 Nonlinear PDE Test Problem: Burgers’ Equation This example demonstrates that WSO avoids order reduction for certain nonlinear IBVPs as well. We consider the viscous Burgers’ equation with pure Neumann b.c. ut + uux = νuxx + f for (x, t) ∈ (0, 1) × (0, 1],

ux = h on {0, 1} × (0, 1] . (7)

Here ν = 0.1 and u(x, t) = cos(2 + 10t) sin(0.2 + 20x). The nonlinear implicit equations arising at each time step are solved using a standard Newton iteration. The choice of Neumann b.c. distinguishes this example from the one given in [9]. With Neumann b.c., the convergence order in u is limited to q˜ + 1.5 (half an order better than with Dirichlet b.c.). Figure 4 shows that order reduction arises with the stage order 1 scheme, and that the WSO 2 scheme recovers 3rd order convergence for u and ux , and the 3rd order WSO 3 scheme yields 3rd order convergence for u, ux and uxx .

4.4 Stiff Nonlinear ODE: Van der Pol Oscillator This example illustrates that DIRK schemes with high WSO may not remove order reduction for all types of nonlinear problems. Consider the Van der Pol oscillator x  = y and y  = μ(1 − x 2 )y − x ,

(8)

with i.c. (x(0), y(0)) = (2, 0), stiffness parameter μ = 500, and final time T = 10. The nonlinear system at each time step is solved via MATLAB’s built-in nonlinear system solver. The “exact” solution is computed using explicit RK4 with a time step Δt = 10−6 . In this case, the presented DIRK schemes with high WSO do not improve the convergence rates in the stiff regime and they perform worse than the WSO 1 scheme in terms of accuracy (see Fig. 5). On the other hand, an EDIRK with

462

D. I. Ketcheson et al. = 500

Van der Pol,

10-8

10-8

10-10

10-10

Error

Error

Van der Pol,

10-12

10-14

= 500

10-12

10-3

10-2 dt

10-1

100

10-14

10-3

10-2 dt

10-1

100

Fig. 5 Error convergence for Van der Pol’s equation. Left: 3rd order DIRK schemes with WSO 1 (blue circles), WSO 2 (red triangles) and WSO 3 (black squares). Right: 4th order DIRK schemes with WSO 1 (blue circles) and WSO 3 (red triangles), and a 3rd order EDIRK scheme with stage order 2 (black squares)

stage order 2 improves the rate of convergence in the stiff regime (see right panel in Fig. 5). However, it does so, interestingly, by yielding larger errors for large time steps.

5 Conclusions and Outlook This study demonstrates that it is possible to overcome order reduction (OR) for certain classes of problems in the context of DIRK schemes, even though these are limited to low stage order. A specific weak stage order (WSO) “eigenvector” criterion has been presented, analyzed, and applied to determine DIRK schemes with WSO up to 3. The numerical results confirm that the schemes avoid OR for linear problems and for some nonlinear problems in which the mechanism for order reduction is linear (i.e., boundary conditions). The key limitation found herein is that the eigenvector criterion cannot go beyond WSO 3 for DIRK schemes. Hence, a key question of future research is how high WSO is admitted by the general criterion in Definition 2. Another important future research task is to devise further DIRK schemes that are truly optimized in terms of truncation error coefficients or other criteria. Acknowledgements This work was supported by the National Science Foundation via grants DMS-1719640 (BS&DZ) and DMS-1719693 (DS); and the Simons Foundation (#359610) (DS).

DIRK Schemes with High Weak Stage Order

463

References 1. Burrage, K., Petzold, L.: On order reduction for Runge-Kutta methods applied to differential/algebraic systems and to stiff systems of ODEs. SIAM J. Numer. Anal. 27(2), 447–456 (1990) 2. Butcher, J.C.: Numerical Methods for Ordinary Differential Equations, 2nd edn. Wiley, New York (2008) 3. Carpenter, M.H., Gottlieb, D., Abarbanel, S., Don, W.-S.: The theoretical accuracy of RungeKutta time discretizations for the initial boundary value problem: a study of the boundary error. SIAM J. Sci. Comput. 16(6), 1241–1252 (1995) 4. Ditkowski, A., Gottlieb, S.: Error inhibiting block one-step schemes for ordinary differential equations. J. Sci. Comput. 73(2–3), 691–711 (2017) 5. Hairer, E., Nørsett, S.P., Wanner, G.: Solving Ordinary Differential Equations I (2nd Revised. Ed.): Nonstiff Problems. Springer, New York (1993) 6. Minion, M.L.: Semi-implicit spectral deferred correction methods for ordinary differential equations. Commun. Math Sci. 1(3), 471–500 (2003) 7. Ostermann, A., Roche, M.: Runge-Kutta methods for partial differential equations and fractional orders of convergence. Math. Comput. 59(200), 403–420 (1992) 8. Prothero, A., Robinson, A.: On the stability and accuracy of one-step methods for solving stiff systems of ordinary differential equations. Math. Comput. 28(125), 145–162 (1974) 9. Rosales, R.R., Seibold, B., Shirokoff, D., Zhou, D.: Order reduction in high-order Runge-Kutta methods for initial boundary value problems (2017). Preprint. arXiv:1712.00897 10. Sanz-Serna, J.M., Verwer, J.G., Hundsdorfer, W.H.: Convergence and order reduction of Runge-Kutta schemes applied to evolutionary problems in partial differential equations. Numer. Math. 50(4), 405–418 (1986) 11. Verwer, J.G.: Convergence and order reduction of diagonally implicit Runge-Kutta schemes in the method of lines. In: Numerical Analysis: Proceedings of the Dundee Conference on Numerical Analysis, 1985, pp. 220–237 (1986) 12. Wanner, G., Hairer, E.: Solving Ordinary Differential Equations II: Stiff and DifferentialAlgebraic Problems, vol. 1. Springer, Berlin (1991)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Scheme for Evolutionary Navier-Stokes-Fourier System with Temperature Dependent Material Properties Based on Spectral/hp Elements Jan Pech

1 Introduction This work presents a numerical algorithm for the system of the Navier-Stokes equations coupled with the balance of internal energy  ρ

  1 ∂v + v · ∇v = −∇p + ∇ · 2μD + λ (∇ · v) I + fv ∂t Re

(1a)

∂ρ + ∇ · (ρv) = m ∂t

(1b)

∂T 1 + v · ∇T = ∇ · (κ∇T ) + fT , ∂t Re Pr

(1c)

where v = [u, v, w]T is the velocity vector (by setting w = const. = 0 we restrict to 2D problem), p is a variable related"to the thermodynamic pressure,1 T denotes ! 1 the temperature, D = 2 ∇v + (∇v)T is the symmetric part of the rate of strain

1 We call thermodynamic pressure the variable acting in the equation of state, e.g. p = ρRT for ideal gas. Quantities with physical units (superscript star) are normalized by its farfield values ∗ ∗ (subscript infinity), e.g. v = |vv∗ | , T = TT ∗ , etc. The dimensionless pressure in (1a) is p = ∗ ρ∞

p∗ . |v∗∞ |2

∞

∞

J. Pech () Institute of Thermomechanics of the Czech Academy of Sciences, Praha 8, Czech Republic e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_37

465

466

J. Pech

tensor, constant Re is the Reynolds number and constant Pr is the Prandtl number (for sake of simplicity we set Re = Pr = 1 for the testing on exact solution). The fluid is expected to be (calorically) perfect,2 Newtonian,3 whose heat flux obeys the Fourier law.4 In system (1), we consider those fluids, which become nonhomogeneous in variable temperature fields due to temperature dependence of its material parameters, namely the density ρ = ρ(T ), dynamic viscosity μ = μ(T ) and thermal conductivity κ = κ(T ). Instead of (1a), we solve  ρ

 ∂v 1 2 + v · ∇v = −∇ p˜ + ∇ · 2μD − μ (∇ · v) I + fv , ∂t Re 3

(2)

where p˜ = p − μb ∇ · v is mean or mechanical pressure, while μb = λ + 23 μ is the bulk viscosity. Equation (2) has the same structure as (1a) while setting λ = − 23 μ (or equivalently μb = 0, c.f. Stokes hypothesis), but physical interpretation of pressure changes. Without loss of generality, solving (2) instead of (1a), we avoid specification of the second viscosity coefficient λ. The forcing terms fv , fT , may represent action of volumetric forces, e.g. gravity or viscous heating, but m is set zero in most of realistic situations. In case of testing of our algorithm on a given solution [ve , pe , Te ]T , we construct the forcing terms such, that Eqs. (2), (1b) and (1c) are satisfied. Our computational scheme is developed for simulations based on the spectral/hp element approximation in spatial coordinates. We use the polynomial approximations of degree 15 in our tests, what eliminates the numerical error in spatial coordinates and we are getting an overview of error production, which belongs directly to the algorithm/discretisation in time. The high order spatial approximations also naturally include approximations of higher-order derivatives, what is utilized in the scheme. The previous results from literature are, up to the authors knowledge, restrictions of (1) setting at least one of the material parameters constant, the velocity field to be divergence-free or modelling a stationary flow, see Table 1.

energy e of the calorically perfect fluids obeys e = cV T , where specific heat at constant volume is independent of temperature (cV = const.). 3 We use the term Newtonian fluid in a general sense for fluids, whose stress tensor is linearly dependent on the strain rate tensor. However, the viscous part of the stress tensor is not traceless as often expected if fluid is called Newtonian. 4 The Fourier law relates the heat flux q to the thermal conductivity κ and the temperature gradient ∇T as q = −κ∇T . 2 Internal

Scheme for Navier-Stokes-Fourier System with Temperature Dependent Properties Table 1 Chosen results concerning equation systems with variable material parameters

[1] [4, 5] [6] [9] [10] [11, 12] [13] [14] [16]

Eq. type nonst. nonst. nonst. stat. nonst. stat. nonst. nonst. nonst.

∇ ·v 0 0 0 0 0 0 = 0 0 0

μ μ(ρ) const. var. μ(T ) μ(T ) μ(T ) const. μ(T ) μ(T )

κ const. const. const. κ(T ) κ(T ) κ(T ) const. κ(T ) κ(T )

467

ρ var. var. const. const. const. const. var. const. const.

Stationary and non-stationary models are denoted stat. and nonst., unspecified variability of a property is denoted var.

2 Algorithm Our approach is inspired by the velocity-correction scheme with the high order pressure boundary condition (HOPBC) proposed for the incompressible NavierStokes equations in [7]. The constant property case, [7], is widely used for its efficiency and was already extended to problems with variable viscosity in [6]. Its modification was used also to the incompressible Navier-Stokes-Fourier system with temperature dependent viscosity and thermal conductivity in [10]. Efficiency of the approach comes from the implicit-explicit (IMEX) formulation, which allows decoupling of the system. The main contribution of the present work, which is a continuation of [10], is in extension to the problems with temperature dependent density. However, the velocity divergence cannot be further neglected in the momentum balance, what is the substantial difference from the previously discussed models and algorithms.

2.1 Decoupled System We use the IMEX scheme in which the Backward difference formula (BDF) of order Q approximates the temporal derivative and a consistent extrapolation is applied to chosen terms (N) F = F u · v ds if F is a

516

G. Nehmetallah et al.

two-dimensional face. Accordingly, for the mesh Th we have (·, ·)Th =



(·, ·)K ,

K∈Th

·, ·Fh =



·, ·∂ Th =

·, ·∂K ,

K∈Th



·, ·F ,

F ∈Fh



·, ·Γa =

·, ·F .

F ∈Fh ∩Γa

We set vt = −n × (n × v) , vn = n (n · v) where vt and vn are the tangential and normal components of v such as v = vt + vn .

3 Principles and Formulation of the HDG Method Following the classical DG approach, approximate solutions (Eh ,Hh ), for all t ∈ [0, T ], are seeked in the space Vh × Vh satisfying for all K in Th ⎧    ⎪ ⎨ ε∂t Eh , v K − curlHh , v K = 0, ∀v ∈ Vh ,     ⎪ ⎩ μ∂t Hh , v + curlEh , v = 0, ∀v ∈ Vh . K K

(5)

Applying Green’s formula, on both equations of (5) introduces boundary terms ˆ h in order to ensure the connection which are replaced by numerical traces Eˆ h and H between element-wise solutions and global consistency of the discretization. This leads to the global formulation for all t ∈ [0, T ] ⎧       ⎪ ˆ h, n × v ⎪ = 0, ∀v ∈ Vh , ⎨ ε∂t Eh , v K − Hh , curlv K + H ∂K       ⎪ ⎪ ⎩ μ∂t Hh , v K + Eh , curlv K − Eˆ h , n × v = 0, ∀v ∈ Vh .

(6)

∂K

It is straightforward to verify that n×v = n×vt and < H, n×v >= − < n×H, v >. ˆt Therefore, using numerical traces defined in terms of the tangential components H h and Eˆ th , we can rewrite (6) as ⎧       ⎪ ˆ t ,n × v ⎪ = 0, ∀v ∈ Vh , ⎨ ε∂t Eh , v K − Hh , curlv K + H h ∂K       ⎪ ⎪ ⎩ μ∂t Hh , v K + Eh , curlv K − Eˆ th , n × v = 0, ∀v ∈ Vh .

(7)

∂K

The hybrid variable Λh introduced in the setting of a HDG method [4] is here defined for all the interfaces of Fh as ˆt, Λh := H h

∀F ∈ Fh .

(8)

An Explicit HDG Method for the 3D Time-Domain Maxwell Equations

517

ˆ t and Eˆ t in each element K of Th by solving We want to determine the fields H h h system (7) and assuming that Λh is known on all the faces of an element K. We consider a numerical trace Eˆ th for all K given by Eˆ th = Eth + τK n × (Λh − Hth ) on ∂K,

(9)

where τK is a local stabilization parameter which is assumed to be strictly positive. We recall that n × Hth = n × Hh . The definitions of the hybrid variable (8) and numerical trace (9) are exactly those adopted in the context of the formulation of HDG methods for the 3D time-harmonic Maxwell equations [10–12, 14]. Following the HDG approach, when the hybrid variable Λh is known for all the faces of the element K, the electromagnetic field can be determined by solving the local system (7) using (8) and (9). From now on we will note by g inc the L2 projection of g inc on Mh . Summing the contributions of (7) over all the elements and enforcing the continuity of the tangential component of Eˆ h , we can formulate a problem which is to find (Eh , Hh , Λh ) ∈ Vh × Vh × Mh such that for all t ∈ [0, T ]       ε∂t Eh , v T − Hh , curlv T + Λh , n × v ∂ T = 0, ∀v ∈ Vh , h h h       μ∂t Hh , v T + Eh , curlv T − Eˆ th , n × v = 0, ∀v ∈ Vh , h ∂Th h     Eˆ h , η − Λh , η Γa − ginc , η = 0, ∀η ∈ Mh , Fh

(10)

Γa

where the last equation is called the conservativity condition with which we ask the tangential component of Eˆ h to be weakly continuous across any interface between two neighboring elements. We now reformulate the system with numerical fluxes. We can deduce from the third equation of (10) that

Λh =

⎧  =  > 1 t t ⎪ 2 τ , H +  E  ⎪ K F h h ⎪ F ⎪ τK + + τK − ⎪ ⎪ ⎨ 1

n × Eth + Hth ,

τK ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ 1

τK + 1

  τK Hth + n × Eth − ginc .

if F ∈ FIh , if F ∈ Fh ∩ Γm ,

(11)

if F ∈ Fh ∩ Γa .

ˆ t,+ = E ˆ t,− with By replacing (11) in (9) we obtain Eˆ th = E h h

Eˆ th =

⎧ #  $  ⎪ 1 t τK + τK − ⎪ t ⎪ E − Hh F , 2 ⎪ ⎪ τK h F ⎪ ⎨ τK + + τK − ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩

0,  1  t Eh − τK n × Hth − τK n × ginc . τK + 1

if F ∈ FIh , if F ∈ Fh ∩ Γm , if F ∈ Fh ∩ Γa .

(12)

518

G. Nehmetallah et al.

Thus, the numerical traces (8) and (9) have been reformulated from the conservativity condition. This means that the conservativity condition is now included in the new formulation of the numerical fluxes and can be neglected in the global system of equations. Hence, the local system (6) takes the form of a classical DG formulation, ∀v ∈ Vh ⎧       ⎪ ˆ t ,n × v ⎪ = 0, ⎨ ε∂t Eh , v K − Hh , curlv K + H h ∂K (13)       ⎪ ⎪ ⎩ μ∂t Hh , v K + Eh , curlv K − Eˆ th , n × v = 0. ∂K

where the numerical fluxes are defined by (11) and (12). √ √ Remark 3 Let YK = εK / μK be the local admittance associated to cell K and ZK = 1/YK the corresponding local impedance. If we set τK = ZK in (11) and 1/τK = YK in (12), the obtained numerical traces coincide with those adopted in the classical upwind flux DGTD method [6].

4 Numerical Results In order to validate and study the numerical convergence of the proposed HDG method, we consider the propagation of an eigenmode in a closed cavity (Ω is the√unit√square) with perfectly metallic walls. The frequency of the wave is f = 3/ 2c0 where c0 is the speed of light in vacuum. The electric permittivity and the magnetic permeability are set to the constant vacuum values. The exact time-domaine solution is given in [6]. We start our study by assuming that the penalization parameter τ is equal to 1. In order to insure the stability of the method, numerical CFL conditions are determined for each value of the interpolation order pK . In our particular case we have εK and μk are constant = 1 ∀K ∈ Th , so we have verified that, as we said in Remark 3, for τ = 1, the values of CFL number correspond to the classical upwind flux-based DG method. In Table 1 we summarize the maximum Δt obtained numerically to insure the stability of the scheme Given these values of Δt max, the L2 -norm of the error is calculated for a uniform tetrahedral mesh with 3072 elements which is constructed from a finite difference grid with nx = ny = nz = 9 points, each cell of this grid yielding 6 tetrahedrons. The wave is propagated in the cavity during a physical time tmax corresponding to 8 periods (as shown in Fig. 1). Figure 2 depicts a comparison of

Table 1 Numerically obtained values of Δt max Interpolation order Δt max (s.)

P1 0.32 × 10−9

P2 0.19 × 10−9

P3 0.13 × 10−9

P4 0.94 × 10−10

Fig. 1 Time evolution of the exact and the numerical solution of Ex at point A(0.25, 0.25, 0.25) with a P3 interpolation

Ex

An Explicit HDG Method for the 3D Time-Domain Maxwell Equations

519

0.4

exact

0.2

numerical

0 −0.2 −0.4

0

1

2 Time(s.)

3 10−8

·10−6

Fig. 2 Time evolution of the L2 -norm of the error for P4 L2 error

4 2 DG- 4 HDG- 4

0 0

1

2

Fig. 3 Numerical convergence order of the time explicit HDG method for τ =1

log(max(L2 error))

Time(s.)

3 10−8

10−1 1

2 3

10−3

1 4 1

10−5

HDG- 1 HDG- 2 HDG- 3

10−1.2

10−1 10−0.8 log(hmin )

10−0.6

the time evolution of the L2 -norm of the error between the solution obtained with an HDG method and a classical upwind flux-based DG method for pK = 4. An optimal convergence with order pK + 1 is obtained as shown in Fig. 3. Now, we keep the same case than previously and we assess the behavior of the HDG method for various values of the penalization parameter τ . We observe that the time evolution of the electromagnetic energy for any order of interpolation, for different values of the parameter τ = 1 and when the Δt used is fixed to the values defined in Table 1, the energy increases in time. In fact, It is necessary to decrease the Δt max for each value of τ to assure the stability (see Table 2 and Fig. 4). For this example, the optimal cost will be for the parameter τ = 1 (having the same cost as an upwind flux for a DG method) otherwise we will spend more time to finish our simulation. On Fig. 5, we show the time evolution of the L2 -error for several values of τ with respect to the maximal time step for the considered parameters. In addition, Table 3 sums up numerical results in term of maximum L2 errors and convergence rates. It appears that the order of convergence is not affected when the stabilization parameter is varied from 1 (with their associated CFL conditions).

520

G. Nehmetallah et al.

Table 2 Numerically obtained values of the CFL number as a function of the stabilization parameter τ for a P1 interpolation τ Δt max (s.)

0.1 0.31×10−10

1.0 3.2×10−10

2.0 1.7×10−10

5.0 0.66×10−10

10.0 0.32×10−10

·10−10  t max(s.)

Fig. 4 Variation of the Δt max as a function of τ

3

1

2

2 3

1 0 0

4



8

6

10

·10−4 L2 error

Fig. 5 Time evolution of the L2 -error as a function of τ with a P3 interpolation

2

0.1 1.0 2.0 5.0 10.0

1.5 1 0.5 0 0

0.5 Time(s.)

1 10−8

Table 3 Maximum L2-errors and convergence orders 1/ h 1/4 1/8 1/16

τ = 1.0 P1 , Δt = 0.16 × 10−09 8.29e−02 – 1.90e−02 2.13 4.74e−03 2.00

P2 , Δt = 0.99 × 10−10 9.87e−03 – 1.34e−03 2.88 1.72e−04 2.97

P3 , Δt = 0.66 × 10−10 9.34e−04 – 5.68e−05 4.04 3.46e−06 4.04

1/ h 1/4 1/8 1/16

τ = 0.1 P1 , Δt = 0.16 × 10−10 2.14e−01 – 5.46e−02 1.97 1.18e−02 2.21

P2 , Δt = 0.96 × 10−11 1.78e−02 – 2.85e−03 2.65 4.06e−04 2.81

P3 , Δt = 0.66 × 10−11 2.19e−03 – 1.68e−04 3.70 1.14e−05 3.88

1/ h 1/4 1/8 1/16

τ = 10.0 P1 , Δt = 0.16 × 10−10 1.74e−01 – 4.24e−02 2.04 9.4e−03 2.16

P2 , Δt = 0.96 × 10−11 1.53e−02 – 2.23e−03 2.76 3.10e−04 2.87

P3 , Δt = 0.68 × 10−11 1.68e−03 – 1.17e−04 3.84 7.81e−06 3.91

An Explicit HDG Method for the 3D Time-Domain Maxwell Equations

521

5 Local Postprocessing We define here, following the ideas of the local postprocessing developed in [1], new n∗ approximations for electric and magnetic field and expect that both En∗ h and Hh n n curl converge with order k + 1 in the H (Th )-norm, whereas Eh and Hh converge curl with order k in the H (Th )-norm. To postprocess Ehn∗ we first compute an n n approximation (p1,h , p2,h ) ∈ V(K) × V(K) to the curl of E, p1 (t n ) = ∇ × E(t n ) and the curl of H, p2 (t n ) = ∇ × H(t n ) by locally solving the below system (pn1,h , v)K = (Enh , ∇ × v)K − Eˆ t,n h , n × v∂K

∀v ∈ V(K)

ˆ t,n , n × v∂K (pn2,h , v)K = (Hnh , ∇ × v)K − H h

∀v ∈ V(K)

and,

n∗ 3 3 We then find (En∗ h , Hh ) ∈ [Pk+1 (K)] × [Pk+1 (K)] such that

⎧ n ⎨ (∇ × En∗ h , ∇ × W)K = (ph,1 , ∇ × W)K , ⎩

n (En∗ h , ∇Y )K = (Eh , ∇Y )K

∀W ∈ [Pk+1 (K)]3 , ∀Y ∈ Pk+2 (K)

and, ⎧ ⎨ (∇ × Hn∗ , ∇ × W)K = (pn , ∇ × W)K , h h,2 ⎩

n (Hn∗ h , ∇Y )K = (Hh , ∇Y )K

∀W ∈ [Pk+1 (K)]3 , ∀Y ∈ Pk+2 (K)

n∗ It is important to point out that we can compute En∗ h and Hh at any time step without advancing in time. Hence, the local postprocessing can be performed whenever we need higher accuracy at particular time steps. Numerical results given in Table 4 shows that a second order convergence rate is obtained for the post-processed solution.

6 Conclusion In this paper we have presented an explicit HDG method to solve the system of Maxwell equations in 3D. The next step is to couple explicit and implicit HDG methods to treat the case of a locally refined mesh.

522 Table 4 Errors and orders of convergence before and after postprocessing

G. Nehmetallah et al.

Pk P1

P2

P3

1/ h 1/4 1/6 1/8 1/4 1/6 1/8 1/4 1/6 1/8

τ = 1.0 ||E − Eh ||Hcurl Error Order 9.30e−01 – 5.84e−01 1.14 4.34e−01 1.03 1.67e−01 – 7.46e−02 1.98 4.29e−02 1.92 2.30e−02 – 7.10e−03 2.90 3.00e−03 2.99

||E − Eh∗ ||Hcurl Error Order 6.83e−01 – 3.10e−01 1.95 1.67e−01 2.15 4.28e−02 – 1.19e−02 3.16 4.90e−03 3.06 5.00e−03 – 1.10e−03 3.79 3.58e−04 3.84

References 1. Abgrall, R., Shu, C.-W.: Handbook of Numerical Methods for Hyperbolic Problems, vol. 17, pp. 190–194. Elsevier/North-Holland, Amsterdam (2016) 2. Carpenter, M.H., Kennedy, C.A.: Fourth-Order 2N-Storage Runge-Kutta Schemes. NASA, Washington (1994) 3. Christophe, A., Descombes, S., Lanteri, S.: An implicit hybridized discontinuous Galerkin method for the 3D time-domain Maxwell equations. Appl. Math. Comput. 319, 395–408 (2018) 4. Cockburn, B., Gopalakrishnan, J., Lazarov, R.: Unified hybridization of discontinuous Galerkin, mixed, and continuous Galerkin methods for second order elliptic problems. SIAM J. Numer. Anal. 47, 1319–1365 (2009) 5. Descombes, S., Lanteri, S. Moya, L.: Locally implicit discontinuous Galerkin time domain method for electromagnetic wave propagation in dispersive media applied to numerical dosimetry in biological tissues. SIAM J. Sci. Comput. 38, A2611–A2633 (2016) 6. Hesthaven, J.S., Warburton, T.: Nodal high-order methods on unstructured grids. I. Timedomain solution of Maxwell’s equations. Int. J. Numer. Methods Eng. 181, 186–221 (2002) 7. Hochbruck, M. Sturm, A.: Error analysis of a second-order locally implicit method for linear Maxwell’s equations. SIAM J. Numer. Anal. 54, 3167–3191 (2016) 8. Kronbichler, M., Schoeder, S., Müller, C., Wall, W.A.: Comparison of implicit and explicit hybridizable discontinuous Galerkin methods for the acoustic wave equation. Int. J. Numer. Methods Eng. 270, 330–342 (2014) 9. Li, L., Lanteri, S. Perrussel, R.: Numerical investigation of a high order hybridizable discontinuous Galerkin method for 2D time-harmonic Maxwell’s equations. COMPEL 2, 1112–1138 (2013) 10. Li, L., Lanteri, S., Perrussel, R.: A hybridizable discontinuous Galerkin method combined to a Schwarz algorithm for the solution of 3D time-harmonic Maxwell’s equations. J. Comput. Phys. 256, 563–581 (2014) 11. Moya, L.: Temporal convergence of a locally implicit discontinuous Galerkin method for Maxwell’s equations. ESAIM Math. Model. Numer. Anal. (M2AN) 46, 1225–1246 (2012) 12. Moya, L., Descombes, S. Lanteri, S.: Locally implicit time integration strategies in a discontinuous Galerkin method for Maxwell’s equations. J. Sci. Comp. 56, 190–218 (2013) 13. Nguyen, N.C., Peraire, J.: Hybridizable discontinuous Galerkin methods for partial differential equations in continuum mechanics. J. Comput. Phys. 231, 5955–5988 (2012) 14. Nguyen, N.C., Peraire, J., Cockburn, B.: Hybridizable discontinuous Galerkin methods for the time-harmonic Maxwell’s equations. J. Comput. Phys. 231, 7151–7175 (2011)

An Explicit HDG Method for the 3D Time-Domain Maxwell Equations

523

15. Stanglmeier, M., Nguyen, N.C., Peraire, J., Cockburn, B.: An explicit hybridizable discontinuous Galerkin method for the acoustic wave equation. Comput. Methods Appl. Mech. Eng. 300, 748–769 (2016) 16. Verwer, J.G.: Component splitting for semi-discrete Maxwell equations. BIT Numer. Math. 51, 427–445 (2011)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Entropy Conserving and Kinetic Energy Preserving Numerical Methods for the Euler Equations Using Summation-by-Parts Operators Hendrik Ranocha

1 Introduction Considering the solution of hyperbolic conservation laws, high order methods can be very efficient, providing accurate numerical solutions with relatively low computational effort [21]. In order to make use of this accuracy, stability has to be established. Mimicking estimates obtained on the continuous level via integrationby-parts, summation-by-parts (SBP) operators [22, 37] can be used. In short, SBP operators are discrete derivative operators equipped with a compatible quadrature providing a discrete analogue of the L2 norm. The compatibility of discrete integration and differentiation mimics integration-by-parts on a discrete level. Combined with the weak enforcement of boundary conditions via simultaneous approximation terms (SATs) [1], highly efficient and stable semidiscretisations can be obtained at least for linear problems, see e.g. [6, 14, 39] and references cited therein. In recent years, there has been an enduring and increasing interest in the basic ideas of SBP operators and their application in various frameworks including finite volume (FV) [25, 26], discontinuous Galerkin (DG) [2, 4, 10, 11, 13, 20, 27, 28, 30], and the recent flux reconstruction/correction procedure via reconstruction framework [15, 16, 42] as described in [31, 32]. While there is only a limited amount of well-posedness theory for nonlinear conservation laws, mimicking properties such as entropy stability semidiscretely has received much interest. Building on the seminal work of Tadmor [40, 41], entropy stability of second order schemes using symmetric numerical fluxes has been investigated, resulting in well-defined properties that numerical fluxes have to satisfy in order to result in entropy conservative

H. Ranocha () TU Braunschweig, Institute Computational Mathematics, Braunschweig, Germany e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_42

525

526

H. Ranocha

schemes. Decomposing general semidiscretisations into a non-dissipative central part and an additional dissipative part, suitable artificial dissipation or filtering can be added afterwards, cf. [7, 9, 38]. Second order methods based on symmetric numerical fluxes can be extended to high order in a conservative way, cf. [4, 7, 28] and [8, 23, 34–36]. Another property of numerical methods for the Euler equations that has received much interest in the literature concerns the kinetic energy. A structural property of numerical fluxes described by Jameson [18] has been used to construct socalled kinetic energy preserving (KEP) numerical fluxes inter alia by Chandrashekar [3]. However, schemes using these fluxes do not preserve the kinetic energy as expected in numerical experiments by Gassner et al. [12]. They had to change the discretisation of the pressure to reduce undesired changes of the kinetic energy. However, this resulted in a loss of entropy conservation. Motivated by these results, some analytical insights into this behaviour have been developed in [29, Section 7.4] and will be presented here. This chapter is structured as follows. At first, some basic results about SBP operators and corresponding semidiscretisations of hyperbolic conservation laws are reviewed in Sect. 2. Afterwards, the Euler equations are considered in Sect. 3. After demonstrating that the property that has been used to characterise numerical fluxes as KEP is not well-defined, the new concept of KEP numerical methods is introduced. Moreover, a numerical flux that is both entropy conservative and kinetic energy preserving in the new sense is developed. Thereafter, results of a numerical experiment comparing entropy conservative numerical fluxes are described in Sect. 4. Finally, a brief summary is given in Sect. 5.

2 Discretisations Using Summation-by-Parts Operators Consider the Euler equations in two space dimensions ⎛

⎛ ⎞ ⎛ ⎞ ⎞ ρvy ρ ρvx ⎜ ⎜ ⎟ ⎜ ⎟ ⎟ ⎜ρv ⎟ ⎜ ρvx2 + p ⎟ ⎜ ρvx vy ⎟ ∂t ⎜ x ⎟ + ∂x ⎜ ⎟ + ∂y ⎜ ⎟ = 0, ⎝ρvy ⎠ ⎝ ρvx vy ⎠ ⎝ ρvy2 + p ⎠ ρe (ρe + p)vx (ρe + p)vy GH I GH I F F F GH I =u

=f x (u)

(1)

=f y (u)

where ρ is the density, v the velocity, e the specific total energy, and p the pressure.  For a perfect gas, p = (γ − 1) ρe − 12 ρv 2 . The usual entropy is U = − γρs −1 , where s = log p − γ log ρ is the specific (physical) entropy. With the entropy fluxes F j fulfilling ∂u U · ∂u f j = ∂u F [19]. The time step Δt has been chosen as Δt = cfl min Δx/(2p + 1)λ , where λ is the greatest absolute value of the eigenvalues of f  and the minimum is taken over all cells and nodes. As in [12], the given numerical fluxes have been used for both the volume terms (7) and as surface fluxes in (6), without additional dissipation. The evolution of the entropy U and the kinetic energy Ekin using a CFL number cfl = 0.9 for the entropy conservative fluxes of Ismail and Roe [17], Chandrashekar [3], and the new flux (11) are visualised in Fig. 1. As can be seen there, the entropy remains approximately constant and the kinetic energy oscillates uniformly until t ≈ 20. Afterwards, the kinetic energy drops for the fluxes of [3, 17] and there is a relative change of the entropy of order 10−5 . Contrary, there is no visible change for the new flux (11). The entropy loss for the fluxes of Ismail and Roe [17] and Chandrashekar [3] is caused by the time integration scheme, as can be seen in Fig. 2, where the time step is reduced by an order of magnitude (cfl = 0.09). However, the behaviour of the kinetic energy is nearly unchanged.

E

(t )− E (0) E (0)

0

− 2 · 10− 2 − 4 · 10− 2

U (t )− U (0) | U (0) |

0

− 1 · 10− 5

Numerical Flux of Chandrashekar Numerical Flux of Ismail & Roe New Numerical Flux (KEP & EC)

− 2 · 10− 5 0

5

10

15 Time t

20

25

30

Fig. 1 Total entropy and kinetic energy of numerical solutions using different entropy conservative numerical fluxes with cfl = 0.9

532

H. Ranocha

Fig. 2 Total entropy and kinetic energy of numerical solutions using different entropy conservative numerical fluxes with cfl = 0.09

5 Summary and Discussion Using summation-by-parts operators, high order numerical schemes with specific properties can be constructed using symmetric (two-point) numerical fluxes. While several “kinetic energy preserving” methods have been proposed, they have been characterised by a property of the numerical fluxes that is not well-defined. Such numerical fluxes resulted in schemes that did not preserve the kinetic energy as expected [12]. Here, a new approach to kinetic energy preservation inspired by the incompressible Euler equations and developed in [29, Section 7.4] has been described. This results in a well-defined property numerical fluxes have to satisfy in order mimic the balance law for the kinetic energy more reliably. Moreover, new entropy conservative numerical fluxes have been developed that are kinetic energy preserving in the new sense.

References 1. Carpenter, M.H., Gottlieb, D., Abarbanel, S.: Time-stable boundary conditions for finitedifference schemes solving hyperbolic systems: methodology and application to high-order compact schemes. J. Comput. Phys. 111(2), 220–236 (1994). https://doi.org/10.1006/jcph. 1994.1057 2. Chan, J.: On discretely entropy conservative and entropy stable discontinuous Galerkin methods. J. Comput. Phys. 362, 346–374 (2018). https://doi.org/10.1016/j.jcp.2018.02.033 3. Chandrashekar, P.: Kinetic energy preserving and entropy stable finite volume schemes for compressible Euler and Navier-Stokes equations. Commun. Comput. Phys. 14(5), 1252–1286 (2013). https://doi.org/10.4208/cicp.170712.010313a

EC and KEP Numerical Methods for the Euler Equations Using SBP Operators

533

4. Chen, T., Shu, C.W.: Entropy stable high order discontinuous Galerkin methods with suitable quadrature rules for hyperbolic conservation laws. J. Comput. Phys. 345, 427–461 (2017). https://doi.org/10.1016/j.jcp.2017.05.025 5. Dafermos, C.M.: Hyperbolic Conservation Laws in Continuum Physics. Springer, Heidelberg (2010). https://doi.org/10.1007/978-3-642-04048-1 6. Fernández, D.C.D.R., Hicken, J.E., Zingg, D.W.: Review of summation-by-parts operators with simultaneous approximation terms for the numerical solution of partial differential equations. Comput. Fluids 95, 171–196 (2014). https://doi.org/10.1016/j.compfluid.2014.02.016 7. Fisher, T.C., Carpenter, M.H.: High-order entropy stable finite difference schemes for nonlinear conservation laws: finite domains. J. Comput. Phys. 252, 518–557 (2013). https://doi.org/10. 1016/j.jcp.2013.06.014 8. Fisher, T.C., Carpenter, M.H., Nordström, J., Yamaleev, N.K., Swanson, C.: Discretely conservative finite-difference formulations for nonlinear conservation laws in split form: theory and boundary conditions. J. Comput. Phys. 234, 353–375 (2013). https://doi.org/10.1016/j.jcp. 2012.09.026 9. Fjordholm, U.S., Mishra, S., Tadmor, E.: Arbitrarily high-order accurate entropy stable essentially nonoscillatory schemes for systems of conservation laws. SIAM J. Numer. Anal. 50(2), 544–573 (2012). https://doi.org/10.1137/110836961 10. Gassner, G.J.: A skew-symmetric discontinuous Galerkin spectral element discretization and its relation to SBP-SAT finite difference methods. SIAM J. Sci. Comput. 35(3), A1233–A1253 (2013). https://doi.org/10.1137/120890144 11. Gassner, G.J.: A kinetic energy preserving nodal discontinuous Galerkin spectral element method. Int. J. Numer. Methods Fluids 76(1), 28–50 (2014). https://doi.org/10.1002/fld.3923 12. Gassner, G.J., Winters, A.R., Kopriva, D.A.: Split form nodal discontinuous Galerkin schemes with summation-by-parts property for the compressible Euler equations. J. Comput. Phys. 327, 39–66 (2016). https://doi.org/10.1016/j.jcp.2016.09.013 13. Gassner, G.J., Winters, A.R., Kopriva, D.A.: A well balanced and entropy conservative discontinuous Galerkin spectral element method for the shallow water equations. Appl. Math. Comput. 272, 291–308 (2016). https://doi.org/10.1016/j.amc.2015.07.014 14. Gustafsson, B., Kreiss, H.O., Oliger, J.: Time-Dependent Problems and Difference Methods. Wiley, Hoboken (2013) 15. Huynh, H.T.: A flux reconstruction approach to high-order schemes including discontinuous Galerkin methods. In: 18th AIAA Computational Fluid Dynamics Conference. American Institute of Aeronautics and Astronautics (2007). https://doi.org/10.2514/6.2007-4079 16. Huynh, H.T., Wang, Z.J., Vincent, P.E.: High-order methods for computational fluid dynamics: a brief review of compact differential formulations on unstructured grids. Comput. Fluids 98, 209–220 (2014). https://doi.org/10.1016/j.compfluid.2013.12.007 17. Ismail, F., Roe, P.L.: Affordable, entropy-consistent Euler flux functions II: entropy production at shocks. J. Comput. Phys. 228(15), 5410–5436 (2009). https://doi.org/10.1016/j.jcp.2009.04. 021 18. Jameson, A.: Formulation of kinetic energy preserving conservative schemes for gas dynamics and direct numerical simulation of one-dimensional viscous compressible flow in a shock tube using entropy and kinetic energy preserving schemes. J. Sci. Comput. 34(2), 188–208 (2008). https://doi.org/10.1007/s10915-007-9172-6 19. Ketcheson, D.I.: Highly efficient strong stability-preserving Runge-Kutta methods with lowstorage implementations. SIAM J. Sci. Comput. 30(4), 2113–2136 (2008). https://doi.org/10. 1137/07070485X 20. Kopriva, D.A., Gassner, G.J.: An energy stable discontinuous Galerkin spectral element discretization for variable coefficient advection problems. SIAM J. Sci. Comput. 36(4), A2076–A2099 (2014). https://doi.org/10.1137/130928650 21. Kreiss, H.O., Oliger, J.: Comparison of accurate methods for the integration of hyperbolic equations. Tellus 24(3), 199–215 (1972). https://doi.org/10.1111/j.2153-3490.1972.tb01547.x

534

H. Ranocha

22. Kreiss, H.O., Scherer, G.: Finite element and finite difference methods for hyperbolic partial differential equations. In: de Boor, C. (ed.) Mathematical Aspects of Finite Elements in Partial Differential Equations, pp. 195–212. Academic, New York (1974) 23. LeFloch, P.G., Mercier, J.M., Rohde, C.: Fully discrete, entropy conservative schemes of arbitrary order. SIAM J. Numer. Anal. 40(5), 1968–1992 (2002). https://doi.org/10.1137/ S003614290240069X 24. Lions, P.L.: Mathematical topics in fluid mechanics. Incompressible Models, vol. 1. Oxford University, Oxford (1996) 25. Nordström, J., Björck, M.: Finite volume approximations and strict stability for hyperbolic problems. Appl. Numer. Math. 38(3), 237–255 (2001). https://doi.org/10.1016/S01689274(01)00027-7 26. Nordström, J., Forsberg, K., Adamsson, C., Eliasson, P.: Finite volume methods, unstructured meshes and strict stability for hyperbolic problems. Appl. Numer. Math. 45(4), 453–473 (2003). https://doi.org/10.1016/S0168-9274(02)00239-8 27. Ortleb, S.: A kinetic energy preserving DG scheme based on Gauss-Legendre points. J. Sci. Comput. 71(3), 1135–1168 (2017). https://doi.org/10.1007/s10915-016-0334-2 28. Ranocha, H.: Comparison of some entropy conservative numerical fluxes for the Euler equations. J. Sci. Comput. (2017). https://doi.org/10.1007/s10915-017-0618-1 29. Ranocha, H.: Generalised summation-by-parts operators and entropy stability of numerical methods for hyperbolic balance laws. Ph.D. Thesis, TU Braunschweig (2018) 30. Ranocha, H.: Generalised summation-by-parts operators and variable coefficients. J. Comput. Phys. 362, 20–48 (2018). https://doi.org/10.1016/j.jcp.2018.02.021 31. Ranocha, H., Öffner, P., Sonar, T.: Summation-by-parts operators for correction procedure via reconstruction. J. Comput. Phys. 311, 299–328 (2016). https://doi.org/10.1016/j.jcp.2016.02. 009 32. Ranocha, H., Öffner, P., Sonar, T.: Summation-by-parts and correction procedure via reconstruction. In: Bittencourt, M.L., Dumont, N.A., Hesthaven, J.S. (eds.) Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2016. Lecture Notes in Computational Science and Engineering, vol. 119, pp. 627–637. Springer, Cham (2017). https://doi.org/10. 1007/978-3-319-65870-4_45 33. Roe, P.L.: Affordable, entropy-consistent Euler flux functions. In: Talk presented at the Eleventh International Conference on Hyperbolic Problems: Theory, Numerics, Applications (2006). http://www2.cscamm.umd.edu/people/faculty/tadmor/references/files/ Roe_Affordable_entropy_Hyp2006.pdf 34. Sjögreen, B., Yee, H.C.: On skew-symmetric splitting and entropy conservation schemes for the Euler equations. In: Kreiss, G., Lötstedt, P., Målqvist, A., Neytcheva, M. (eds.) Numerical Mathematics and Advanced Applications 2009: Proceedings of ENUMATH 2009, the 8th European Conference on Numerical Mathematics and Advanced Applications, Uppsala, July 2009, pp. 817–827. Springer, Heidelberg (2010). https://doi.org/10.1007/978-3-642-11795-4_ 88 35. Sjögreen, B., Yee, H.: High order entropy conservative central schemes for wide ranges of compressible gas dynamics and MHD flows. J. Comput. Phys. 364, 153–185 (2018). https:// doi.org/10.1016/j.jcp.2018.02.003 36. Sjögreen, B., Yee, H.C., Kotov, D.: Skew-symmetric splitting and stability of high order central schemes. In: Journal of Physics: Conference Series, vol. 837, p. 012019. IOP Publishing, Philadelphia (2017). https://doi.org/10.1088/1742-6596/837/1/012019 37. Strand, B.: Summation by parts for finite difference approximations for d/dx. J. Comput. Phys. 110(1), 47–67 (1994). https://doi.org/10.1006/jcph.1994.1005 38. Svärd, M., Mishra, S.: Shock capturing artificial dissipation for high-order finite difference schemes. J. Sci. Comput. 39(3), 454–484 (2009). https://doi.org/10.1007/s10915-009-9285-1 39. Svärd, M., Nordström, J.: Review of summation-by-parts schemes for initial-boundary-value problems. J. Comput. Phys. 268, 17–38 (2014). https://doi.org/10.1016/j.jcp.2014.02.031

EC and KEP Numerical Methods for the Euler Equations Using SBP Operators

535

40. Tadmor, E.: The numerical viscosity of entropy stable schemes for systems of conservation laws. I. Math. Comput. 49(179), 91–103 (1987). https://doi.org/10.1090/S0025-5718-19870890255-3 41. Tadmor, E.: Entropy stability theory for difference approximations of nonlinear conservation laws and related time-dependent problems. Acta Numer. 12, 451–512 (2003). https://doi.org/ 10.1017/S0962492902000156 42. Wang, Z.J., Gao, H.: A unifying lifting collocation penalty formulation including the discontinuous Galerkin, spectral volume/difference methods for conservation laws on mixed grids. J. Comput. Phys. 228(21), 8161–8186 (2009). https://doi.org/10.1016/j.jcp.2009.07.036

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Multiwavelet Troubled-Cell Indication: A Comparison of Utilizing Theory Versus Outlier Detection Mathea J. Vuik

1 Introduction Solutions to nonlinear hyperbolic PDEs develop discontinuities in time. The generation of spurious oscillations in such regions can be prevented by applying a limiter in the troubled zones. In [16, 18], two different multiwavelet troubledcell indicators were introduced, one based on a parameter, the other using outlier detection. We present this comparison in order to begin to understand in which regime these tools are effective. In this paper, we investigate the effectiveness of a different detection scheme, based on the theoretical detection of troubled cells using multiwavelet approaches. It uses the cancelation property [6] and the theory about thresholding [8]. This technique was originally used for a multiwaveletbased adaptive strategy in combination with the DG method. However, we are specifically interested in its application for troubled-cell indication. In the troubled cells, the moment limiter is applied [11]. We demonstrate the performance of this new indicator and show that it works very well when very fine meshes are used (the asymptotic regime). For coarser meshes, it seems that the existing multiwavelet troubled-cell indicators perform better. The outline of this paper is as follows: in Sect. 2, some background information about the multiwavelet theory is given. The existing multiwavelet troubled-cell indicators, as well as the cancelation property and the derived thresholding technique are described in Sect. 3. Numerical results are shown in Sect. 4, and some concluding remarks are given in Sect. 5.

M. J. Vuik () VORtech, Delft, The Netherlands e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_43

537

538

M. J. Vuik

2 Multiwavelets and DG In this section, we consider the multiwavelet theory that is used to design the different troubled-cell indicators. For the sake of brevity, we neglect discussion of the DG scheme [4, 5], that is used in the computations. The relation between the DG scheme and multiwavelets was shown in [16]. Any global one-dimensional DG approximation of degree k can be written as uh (x) = 2

− n2

n −1 k 2 

()

n uj φj (x),

j =0 =0 n are the scaling functions related to the orthonormal Legendre polynomiwhere φj als. The corresponding multiwavelet decomposition is

uh (x) =

k  =0

n−1 2 −1  k  m

0 s0 φ (x) +

m m dj ψj (x),

m=0 j =0 =0

0 are the scaling-function coefficients belonging to u , and d m are the where s0 h j corresponding multiwavelet coefficients, [2, 16]. The multiwavelets ψ have been developed by Alpert [1].

3 Utilizing Multiwavelet Coefficients for Troubled-Cell Indication In this section, we show different troubled-cell indicators that utilize multiwavelet coefficients. Note that, as the detectors are solely based on the underlying approximation space, the ideas do not need to be modified in order to be applied to other types of model problems than those included in this paper. First, the existing indicators that use either a parameter or the boxplot method are presented. Next, the cancelation property and thresholding technique are used to design a different indication technique.

3.1 Boxplots for Outlier Detection n−1 In [16, 17], we have shown that the coefficients dkj are very useful for troubledcell indication. With this knowledge, we have designed two different troubledcell indicators. The first indicator is the so-called parameter-based multiwavelet

Multiwavelet Troubled-Cell Indication Comparisons

539

troubled-cell indicator [16]. Here, we detect an element as troubled when n−1 n−1 |dkj | > C · max{|dkj |, j = 0, . . . , 2n − 1}, C ∈ [0, 1].

(1)

The value of C is a useful tool to prescribe the strictness of the limiter. n−1 Another option is to use outlier detection on the multiwavelet coefficients dkj to detect the troubled cells [18]. Here, Tukey’s boxplot method [14] is applied locally to prevent the need for a problem-dependent parameter. The different steps are presented in Algorithm 1. Algorithm 1 Outlier-detection algorithm using local vectors Send in a suitable troubled-cell indication vector D. Split this vector into local vectors, d. for all local vectors do Sort d to obtain ds . Compute the quartiles Q1 and Q3 . Detect djs in the smallest 25% of ds if djs < Q1 − 3(Q3 − Q1 ), and djs in the biggest 25% of ds if djs > Q3 + 3(Q3 − Q1 ). end for Ignore the detected outliers in the left half of the local region when they are not detected with respect to the left-neighboring vector, and similarly test the detected coefficients in the right half of the local region.

Outliers are the coefficients in the vector that are straying far out beyond the others. In order to pick out certain coefficients as outliers, the outer fences are constructed, which were originally defined by Tukey [14]. The outer fences of a vector are [Q1 − 3(Q3 − Q1 ), Q3 + 3(Q3 − Q1 )] (coefficients outside are called extreme outliers). The coverage for this whisker length is 99.9998%, such that only 0.0002% of the data in a normally distributed vector is detected as an extreme outlier (asymptotically) [9]. In our computations, we always use local vectors of length 16.

3.2 Cancelation Property In this section, the cancelation property is stated and proved for the one-dimensional case [6]. Here, we assume that the multiwavelets have M + 1 vanishing moments. In our case, we have M =  + k [1, 15]. If the solution satisfies the continuity requirement u|Ijm ∈ C M+1 (Ijm ) (where Ijm is the j -th element in level m), then m dj ≤

1 · ||u(M+1) ||L∞ (Ijm ) · 2(−m+1)(M+3/2), (M + 1)!

m = 0, . . . , n, j = 0, . . . , 2m − 1,  = 0, . . . , k.

(2)

540

M. J. Vuik

The proof uses a Taylor expansion of u about element center xjm : there exists a ξ between x and xjm such that u(x) = u(xjm ) + u (xjm )(x − xjm ) + . . . +

u(M) (xjm ) M!

(x − xjm )M +

u(M+1) (ξ ) (x − xjm )M+1 . (M + 1)!

Using that the first M + 1 moments of the multiwavelets vanish, we find C

m m m dj = u, ψj Ij

≤

u(M+1) (ξ ) m (x − xjm )M+1 , ψj = (M + 1)!

D Ijm

1 m m Ij . ||u(M+1) ||L∞ (Ijm ) (x − xjm )M+1 , ψj (M + 1)!

(3)

Next, we use Cauchy-Schwarz’s inequality to find m  m ≤ ||(x − x m )M+1 || m m M+1 || (x − xjm )M+1 , ψj I L2 (I m ) · ||ψj ||L2 (I m ) = ||(x − xj ) L2 (I m ) , j j

j

j

j

because the multiwavelets are orthonormal. Using the notation Δx m for the element size in level m, we have √ ||(x − xjm )M+1 ||L2 (I m ) ≤ (Δx m )M+1 ||1||L2 (I m ) = (Δx m )M+1 Δx m = (Δx m )M+3/2 . j

j

For the domain [−1, 1], we have Δx m = 2−m+1 . This means that ||(x − xjm )M+1 ||L2 (I m ) ≤ 2(−m+1)(M+3/2), j

which proves the cancelation property. It should be noticed that this result can be generalized to general grid hierarchies and higher-dimensional problems [6, 10]. The next section contains a discussion of the thresholding technique for onedimensional multiwavelet expansions.

3.3 Thresholding of the Multiwavelet Coefficients In this section, the thresholding technique for systems of conservation laws in one dimension is explained, which is based on the cancelation property [8]. This technique is originally used for a multiwavelet-based adaptive strategy in combination with the DG method. However, we are specifically interested in its application for troubled-cell indication.

Multiwavelet Troubled-Cell Indication Comparisons

541

Following [8], the element Ijn−1 is detected as troubled if ⎛ ⎜ max ⎝

=0,...,k r=1,2,3

⎞ n−1 |dj (r)|

√ ⎟   ⎠ > εn−1 2Δx. n (r)|, 1 max maxj =0,...,2n −1 2(n−1)/2|s0j

Here, the value √ r is related to the conserved quantity in a system of three PDEs. The factor 2Δx (with Δx the DG mesh width) occurs because of a scaling difference: the multiwavelets in [8] are scaled with respect to the L∞ -norm, whereas an L2 -norm scaling is used in this paper. The level-dependent threshold value εn−1 is chosen as εn−1 = ε/2. The parameter ε can be chosen using two different strategies [8]. The first option is to use the a priori strategy, which is based on the balance between discretization errors and perturbation errors of adaptive meshes [10]. If the solution contains discontinuities, then the a priori strategy leads to ε = CΔx 2 . The second option is the heuristic approach, which is based on numerous computations for practical applications [8]. This method is more efficient since it is less pessimistic than the a priori strategy. For discontinuous solutions, the heuristic approach uses ε = CΔx. This yields detection of element Ijn−1 if ⎞

⎛ ⎜ max ⎝

=0,...,k r=1,2,3

n−1 (r)| |dj

1 ⎟   ⎠ > √ Δx β+0.5 C, n (n−1)/2 2 max maxj =0,...,2n −1 2 |s0j (r)|, 1

where β = 2 for the a priori strategy and β = 1 for the heuristic strategy. Note that the multiwavelet coefficients are scaled by the cell average if this value is greater than 1 in absolute value (to prevent division by zero). The optimal choice of the parameter C depends on the problem, in particular on the strength of the shock compared to the normal amplitude of the solution. The smaller C is, the more elements are detected. In general, the value C = 1/(b − a) should work for the domain [a, b] [8]. If C is chosen too small, then too many cells are detected as troubled. For the adaptive strategy, this is not really problematic since the approximation is usually more accurate on a finer grid. However, for troubledcell indication, it is important to detect the correct number of elements. It should be noticed that this indicator is designed for very fine resolutions (since the strategies use asymptotic arguments). For coarse meshes, smaller values of C should be used, which are difficult to predict a priori.

542

M. J. Vuik

3.4 Generalized Grids The algorithm for utilizing Alpert’s multiwavelets for a nonuniform grid is given in [7]: the only difference with Alpert’s algorithm [1] is that no additional vanishing moments are added. Multiwavelets for one-dimensional irregular meshes have been designed in [12, 13]. It should be noticed that this construction is local, which means that the resulting bases are depending on the level and the position unless there is an affine mapping from the element to a reference element. This leads to slower computations. On the other hand, the use of such multiwavelet space makes it possible to decompose the DG approximation to a multiwavelet expansion exactly. The multiwavelet coefficients will again become small if the underlying function is smooth, and the mesh width between two neighboring elements is not varying too much. When coupled with a troubled-cell indication variable, it will be necessary to include spatial information of the mesh in the algorithm using the element size. Alternatively, one can use of a window-based technique [3]. A window is a fixed length subsequence of the test sequence, which can be slid through the domain using a sliding step. These issues and resulting numerics are discussed further in [15].

4 Numerical Results In this section, the different multiwavelet troubled-cell indicators are applied to onedimensional problems based on the Euler equations of gas dynamics. The results for the original multiwavelet troubled-cell indicators (both based on a parameter, and based on outlier detection), can be seen in Figs. 1 and 2 (polynomial degree 2, 128 elements for Sod’s and Lax’s shock tube, and 512 elements for the blast-wave and Shu-Osher problem). The parameter-based technique performs well if a suitable value for the problem-dependent parameter C is chosen. The outlierdetection results are generally better than the original troubled-cell indicator using an optimized parameter: both the weak and the strong shock regions were detected, whereas smooth regions were not selected. It is also possible to use the thresholding technique for multiwavelet coefficients to detect troubled cells. It turns out that this indicator works very well as long as an appropriate value for C is chosen, and the mesh is taken fine enough. The results for the different test cases are visualized in Fig. 3 using the heuristic strategy (polynomial degree 2, 1024 elements for all models). Here, we take the value C = 1/(b − a) where [a, b] is the domain on which the test problem is defined. Note that this thresholding technique is very accurate. However, many elements should be used to meet the asymptotic properties of the indicator.

Multiwavelet Troubled-Cell Indication Comparisons

543

2 1.2

1.8 1.6

1

1.4 0.8

1

t

t

1.2

0.6

0.8 0.4

0.6 0.4

0.2

0.2 −5

−4

−3

−2

−1

0 x

1

2

3

4

5

−5

−4

−3

−2

−1

(a)

0 x

1

2

3

4

5

(b) 1.8 1.6

0.03

1.4

0.025

1.2 1

0.02

t

t

0.035

0.8

0.015

0.6 0.01

0.4

0.005

0.2 0

0.2

0.6

0.4 x

(c)

0.8

1

−5

−4

−3

−2

−1

0 x

1

2

3

4

5

(d)

Fig. 1 Time-history plot of detected troubled cells using the parameter-based multiwavelet troubled-cell indicator, polynomial degree 2. (a) Sod’s shock tube, C = 0.1, 128 elements. (b) Lax’s shock tube, C = 0.1, 128 elements. (c) Blast-wave problem, C = 0.05, 512 elements. (d) Shu-Osher, C = 0.01, 512 elements

If the number of elements is taken smaller, then C should decrease to detect the correct features. In that case, it is difficult to guess the correct value of C. Another option is to use the a priori strategy for coarser meshes, see Fig. 4 (polynomial degree 2, 128 elements for Sod’s and Lax’s shock tube, and 512 elements for the blast-wave and Shu-Osher problem). If C = 1/(b − a) is used, then this approach works well for Sod’s and Lax’s shock tube, but too many elements are detected for the blast-wave and the Shu-Osher problem. Also here, the value of C should be adapted to find the correct results.

544

M. J. Vuik 2 1.2

1.8 1.6

1

1.4 0.8

1

t

t

1.2

0.6

0.8 0.4

0.6 0.4

0.2

0.2 −5

−4

−3

−2

−1

0 x

1

2

3

4

5

−5

−4

−3

−2

−1

(a)

0 x

1

2

3

4

5

(b) 1.8 1.6

0.03

1.4

0.025

1.2 1

0.02

t

t

0.035

0.8

0.015

0.6 0.01

0.4

0.005 0

0.2 0.2

0.4

0.6

0.8

1

−5

−4

−3

−2

−1

x

0 x

(c)

(d)

1

2

3

4

5

Fig. 2 Time-history plot of detected troubled cells using the outlier-detection multiwavelet troubled-cell indicator, polynomial degree 2. (a) Sod’s shock tube, 128 elements. (b) Lax’s shock tube, 128 elements. (c) Blast-wave problem, 512 elements. (d) Shu-Osher problem, 512 elements

5 Conclusions and Recommendations In this paper, a new troubled-cell indicator was formed, based on the cancelation property for multiwavelets and the derived thresholding technique. Inspection of this technique reveals that it is very useful to design adaptive meshes [8]. For troubled-cell indication, we found out that detection is very accurate as long as a very fine mesh is used. For coarser meshes, it seems to be more useful to apply a different detection method. Furthermore, it is not straightforward how to choose the parameter C. More research should be done to see in which way the cancelation property for multiwavelet coefficients can be used for the accurate detection of troubled cells. For example, it could be that this property also relates to the severity of the shocks.

Multiwavelet Troubled-Cell Indication Comparisons

545

2 1.2

1.8 1.6

1

1.4 0.8

1

t

t

1.2

0.6

0.8 0.4

0.6 0.4

0.2

0.2 −5

−4

−3

−2

−1

0 x

1

2

3

4

−5

5

−4

−3

−2

−1

0 x

1

2

3

4

5

(b)

(a) 1.8 1.6

0.03

1.4

0.025

1.2 1

0.02

t

t

0.035

0.8

0.015

0.6 0.01

0.4

0.005 0

0.2 0.2

0.4

0.6 x

(c)

0.8

1

−5

−4

−3

−2

−1

0 x

1

2

3

4

5

(d)

Fig. 3 Thresholding technique with heuristic approach: time-history plot of detected troubled cells, 1024 elements, polynomial degree 2, C = 1/(b − a), with [a, b] the computational domain. (a) Sod’s shock tube. (b) Lax’s shock tube. (c) Blast-wave problem. (d) Shu-Osher problem

546

M. J. Vuik 2 1.2

1.8 1.6

1

1.4 0.8

1

t

t

1.2

0.6

0.8 0.4

0.6 0.4

0.2

0.2 −5

−4

−3

−2

−1

0 x

1

2

3

4

−5

5

−4

−3

−2

−1

0 x

1

2

3

4

5

(b)

(a) 1.8 0.035

1.6

0.03

1.4

0.025

1.2 1 t

t

0.02

0.8

0.015

0.6 0.01

0.4

0.005

0.2 0

0.2

0.4

0.6 x

(c)

0.8

1

−5

−4

−3

−2

−1

0 x

1

2

3

4

5

(d)

Fig. 4 Thresholding technique with a priori approach on coarser meshes: time-history plot of detected troubled cells, polynomial degree 2, C = 1/(b−a), with [a, b] the computational domain. (a) Sod’s shock tube, 128 elements. (b) Lax’s shock tube, 128 elements. (c) Blast-wave problem, 512 elements. (d) Shu-Osher problem, 512 elements

Multiwavelet Troubled-Cell Indication Comparisons

547

Acknowledgements The author gratefully wishes to acknowledge the collaboration with Jennifer Ryan and the useful comments provided by Siegfried Müller that helped to shape this work.

References 1. Alpert, B.K.: A class of bases in L2 for the sparse representation of integral operators. SIAM J. Math. Anal. 24, 246–262 (1993) 2. Archibald, R.K., Fann, G.I., Shelton, W.A.: Adaptive discontinuous Galerkin methods in multiwavelets bases. Appl. Numer. Math. 61, 879–890 (2011) 3. Chandola, V.: Anomaly detection for symbolic sequences and time series data. PhD Thesis, University of Minnesota, Minneapolis (2009) 4. Cockburn, B., Shu, C.-W.: TVB Runge-Kutta local projection discontinuous Galerkin finite element method for conservation laws II: general framework. Math. Comput. 52, 411–435 (1989) 5. Cockburn, B., Lin, S.-Y., Shu, C.-W.: TVB Runge-Kutta local projection discontinuous Galerkin finite element method for conservation laws III: one-dimensional systems. J. Comput. Phys. 84, 90–113 (1989) 6. Dahmen, W.: Wavelet methods for PDEs—some recent developments. J. Comput. Appl. Math. 128, 133–185 (2001) 7. Gerhard, N., Müller, S.: Adaptive multiresolution discontinuous Galerkin schemes for conservation laws: multi-dimensional case. Comput. Appl. Math. 35, 321–349 (2016) 8. Gerhard, N., Iacono, F., May, G., Müller, S., Schäfer, R.: A high-order discontinuous Galerkin discretization with multiwavelet-based grid adaptation for compressible flows. J. Sci. Comput. 62, 25–52 (2015) 9. Hoaglin, D.C., Iglewicz, B., Tukey, J.W.: Performance of some resistant rules for outlier labeling. J. Am. Statist. Assoc. 81, 991—999 (1986) 10. Hovhannisyan, N., Müller, S., Schäfer, R.: Adaptive multiresolution discontinuous Galerkin schemes for conservation laws. Math. Comput. 83, 113–151 (2014) 11. Krivodonova, L.: Limiters for high-order discontinuous Galerkin methods. J. Comput. Phys. 226, 879–896 (2007) 12. Nagel, D.: Effiziente Konstruktion von Multiwavelets auf nicht uniformen dyadischen Gitterhierarchien. MasterÕs thesis from RWTH Aachen University (2015) 13. Pistre, S.: Konstruktion von Multiwavelets auf nicht-uniformen eindimensionalen Gitterhierarchien. BachelorÕs thesis from RWTH Aachen University (2013) 14. Tukey, J.W.: Exploratory Data Analysis. Addison-Wesley, Boston (1977) 15. Vuik, M.J.: The use of multiwavelets and outlier detection for troubled-cell indication in discontinuous Galerkin methods. PhD Thesis from Delft University of Technology (2017) 16. Vuik, M.J., Ryan, J.K.: Multiwavelet troubled-cell indicator for discontinuity detection of discontinuous Galerkin schemes. J. Comput. Phys. 270, 138–160 (2014) 17. Vuik, M.J., Ryan, J.K.: Multiwavelets and Jumps in DG Approximations. In: Kirby, R.M., Berzins, M., Hesthaven, J.S. (eds.) Spectral and High Order Methods for Partial Differential Equations—ICOSAHOM 2014, pp. 503–511. Springer, Berlin (2015) 18. Vuik, M.J., Ryan, J.K.: Automated parameters for troubled-cell indicators using outlier detection. SIAM J. Sci. Comput. 38, A84–A104 (2016)

548

M. J. Vuik

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

An Anisotropic p-Adaptation Multigrid Scheme for Discontinuous Galerkin Methods Andrés M. Rueda-Ramírez, Gonzalo Rubio, Esteban Ferrer, and Eusebio Valero

1 Introduction In recent decades, high-order discontinuous Galerkin (DG) methods have been gaining increasing popularity for high-accuracy solutions of systems of conservation laws, such as the compressible Euler and Navier-Stokes equations [5, 6, 22]. The lack of a continuity constraint on element interfaces makes DG methods robust for describing advection-dominated problems when an appropriate Riemann solver is selected [5, 12, 22]. Multigrid methods speed up the iterative solution of large systems of equations using coarse-grid representations (lower levels). Iterative methods (known as smoothers in the multigrid community) are good at eliminating the high frequencies of the error fast; therefore, when applied to coarse-grid representations, they also reduce the low frequencies of the error. They have been broadly used in the highorder community in recent years in the form of p-multigrid [2, 8] (where levels are constructed using different polynomial orders) and hp-multigrid [14, 21] (where both the order and size of the elements are changed). Two types of multigrid methods can be found in the literature: linear and nonlinear multigrid. In our work, we make use of the nonlinear multigrid scheme, also known as the Full Approximation Scheme (FAS), since it enables the estimation of the truncation error of coarse representations, as will be shown. The smoother can be either a time-marching scheme (implicit or explicit), or an iterative method applied to the linearized problem.

A. M. Rueda-Ramírez () · G. Rubio · E. Ferrer · E. Valero ETSIAE-UPM (School of Aeronautics - Universidad Politécnica de Madrid), Plaza Cardenal Cisneros, Madrid, Spain e-mail: [email protected]; [email protected]; [email protected]; [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_44

549

550

A. M. Rueda-Ramírez et al.

Because of the allowed discontinuities on element interfaces, DG methods are capable of handling non-conforming meshes with hanging nodes and/or different polynomial orders efficiently [7, 13, 15]. It is possible to take advantage of this feature to accelerate the computations through local adaptation strategies. Local adaptation can be performed by subdividing or merging elements (h-adaptation) or by enriching or reducing the polynomial order in certain elements (p-adaptation). The main idea behind these methodologies is to reduce the number of degrees of freedom (NDOF) while maintaining a high accuracy, which translates into shorter computational times and reduced storage requirements. Furthermore, since several 2D and 3D implementations of the DG methods use tensor-product basis functions, it is possible to adapt the polynomial order in each coordinate direction independently. In order to identify the localized regions that need increased or decreased accuracy, an error estimator is commonly used. There are several approaches to estimate the error and drive an adaptation method. In this work, we focus on truncation error estimates since it has been shown that a reduction of the truncation error controls the numerical accuracy of all functionals [10], hence reducing the truncation error necessarily leads to a more accurate lift and drag. The τ -estimation method [4] is a way to estimate the truncation error locally that has been used to drive mesh adaptation strategies in low-order [9, 20] and high-order methods [10, 17, 18]. The adaptation strategy consists in converging a high order representation (reference mesh) to a specified global residual and then performing a single error estimation followed by a corresponding mesh adaptation process. Rueda-Ramírez et al. [19] developed a new method for estimating the truncation error of anisotropic representations that is cheaper to evaluate than previous implementations, and showed that it produces very accurate extrapolations of the truncation error, which enables the use of coarser reference meshes. In this work, we employ the anisotropic truncation error estimator developed in [19] and the anisotropic p-adaptation method detailed in [18] to accelerate the computation of the compressible steady viscous flow past a NACA0012 at angle of attack 5◦ , Re∞ = 200 based on the airfoil chord, and M∞ = 0.2. This particular settings correspond to a steady laminar flow, but the proposed method can be directly used with any steady solution (e.g. RANS). The paper is organized as follows: In Sect. 2, we briefly describe the methods used in this paper. In Sect. 3, we compare the performance of the proposed methods with traditional strategies for solving the flow past a NACA0012 and show the speed-up advantages for different accuracies. Finally, the conclusions are summarized in Sect. 4.

2 Methods 2.1 DG Method We consider the approximation of systems of conservation laws, ∂t q + ∇ · F = s,

(1)

An Anisotropic p-Adaptation Multigrid Scheme for the DGSEM

551

where q is the vector of conserved variables, F is the flux dyadic tensor, and s is a source term. The domain Ω is partitioned in a mesh T = {e} consisting of K non-overlapping elements Ω e . Multiplying equation (1) by a test function v and integrating by parts over each subdomain Ω e yields the weak formulation: 

 Ωe

∂t qvdΩ e −

 Ωe

F · ∇vdΩ e +

 ∂Ω e

F · nvdσ e =

svdΩ e .

(2)

Ωe

Let q, s, F and v be approximated by piece-wise polynomial functions defined in the space of L2 functions: V N = {vN ∈ L2 (Ω e ) : vN |Ω e ∈ PN (Ω e ) ∀ Ω e ∈ T }, where PN (Ω e ) is the space of polynomials of degree at most N. The functions in VN can be represented in each element as a linear combination of basis functions < N N φiN ∈ PN (Ω e ) (e.g. qN |Ω e = i QN φ i i ), where φi are usually tensor product expansions. After some manipulations, the discontinuous Galerkin finite element discretization system is obtained: [M]∂t QN + F(QN ) = [M]SN ,

(3)

where [M] is the mass matrix and F is a nonlinear operator, which are the assembled global versions of the element-wise mass matrices and nonlinear operators:  [M]ei,j = F (Q)j = e

Ωe

φi φj dΩ e ,

e NDOF  

i=1

 −

Ωe

(4) 

F ei · φi ∇φj dΩ e +

∂Ω e

  F∗ N Q, Q− , n φj dσ e , (5)

where F ei is the ith position of the vector F e , which contains the value of Fe for all the degrees of freedom of element e. In the rest of this paper, bold uppercase Roman letters and bold Greek letters are used to note vectors spanning several degrees of freedom, unless specified. The numerical flux function F∗ allows to uniquely define the flux at the element interfaces and to weakly prescribe the boundary data as a function of the conserved variable on both sides of the boundary/interface and the normal vector. In the present work, we use the scheme by Roe [16] as the advective Riemann solver and the original scheme by Bassi and Rebay [1] (BR1) as the diffusive Riemann solver.

2.2 Full Approximation Scheme p-Multigrid The Full Approximation Scheme (FAS) is a nonlinear version of the multigrid method that is specially suited to solve systems of nonlinear equations [4]. Departing from Eq. (3) and defining the operator A(QN ) = [M]−1 F(QN ), the steady-state

552

A. M. Rueda-Ramírez et al.

problem of order P yields A(QP ) = SP .

(6)

˜ P is obtained that has After β1 sweeps of a smoother, a non-converged solution Q P P P ˜ an associated discretization error # = Q − Q . The FAS multigrid procedure consists in obtaining an approximation to the discretization error in a coarse grid of order N and projecting it to the original problem of order P : ˜P # P = IPN # N = IPN (QN − IN P Q ),

(7)

where IPN is an L2 projection operator N → P and QN is the solution to the coarsegrid problem: AN (QN ) = SN ,

(8)

where the source term is defined as   N P P ˜P ˜P SN = AN (IN P Q ) + IP S − A (Q ) .

(9)

In practice, several p-multigrid levels are used in V- or W-cycles. The smoothing steps that are performed when coarsening are called pre-smoothing sweeps, and the ones performed when refining back are called post-smoothing sweeps. Furthermore, QN is not obtained exactly in the coarse grids, but approximated using an iterative ˜ N → QN . In this work, we use a third order low-storage Runge-Kutta method Q (RK3) as the smoother and V-cycles.

2.3 τ -Based p-Adaptation In this section we show how to drive an anisotropic p-adaptation procedure using the truncation error, which is estimated in the multigrid procedure.

2.3.1 The Anisotropic τ -Estimation Method The non-isolated truncation error of a discretization of order N is defined as τ N = RN (IN q) − R(q),

(10)

where q is the exact solution to the problem, IN is a discretizing operator, R is the continuous partial differentiation operator, and RN is the discrete partial

An Anisotropic p-Adaptation Multigrid Scheme for the DGSEM

553

differentiation operator. From Eqs. (1) and (3): R(q) = s − ∇ · F,

(11)

R N (II N q) = [M]SN − F(II N q),

(12)

where I N is an operator that samples the exact solution on the points that correspond to the degrees of freedom of a representation of order N, and therefore Eq. (12) corresponds to the sampled values of RN (IN q). Note that in steady cases, R(q) = 0 holds. Since the exact solution q is usually not at hand, we utilize the quasi a-piori τ -estimation method, which approximates the exact solution with the non-converged solution on a high-order grid q ≈ q˜ P , where N < P . Therefore, the steady non-isolated truncation error estimation yields N N ˜P N N ˜P ˜P ) → τ N τPN = RN (IN Pq P = R (IP Q ) = [M]S − F(IP Q ).

(13)

On the left side of the arrow is the estimation of the truncation error that lives in the space VN , and on the right side is the sampled form of the truncation error estimation on the points that correspond to the degrees of freedom. In a DG representation, one can also define the isolated truncation error τˆ as N ˆ N ˜P ˆN N ˜P τˆ N P = R (IP Q ) = [M]S − F(IP Q ),

(14)

where Fˆ is the assembled version of the isolated nonlinear operator, defined elementwise as Fe (Q)j =

e NDOF  

i=1

 −

Ωe

 F ei · φi ∇φj dΩ e +

∂Ω e

FN · nφj dσ e .

(15)

Note that Eq. (15) is (5) without substituting F by the numerical flux F∗ . This change eliminates the influence of the neighboring elements and boundaries on the truncation error of each element. We drop the hat notation in the next statements since they are valid for both the isolated and non-isolated truncation error. The τ -estimation method can also be used with anisotropic representations, i.e. τPN11PN22 = RN1 N2 (IPN11PN22 q˜ P1 P2 ),

(16)

where Ni and Pi are the polynomial orders in the direction i of the analyzed representation and the high-order reference solution, respectively, where Ni < Pi . Additionally, Rueda-Ramírez et al. [19] showed that the truncation error of an anisotropic representation can be estimated using directional components: τ N1 N2 ≈ τ1N1 N2 + τ2N1 N2 ≈ τPN11PP22 + τPP11PN22 ,

(17)

554

A. M. Rueda-Ramírez et al.

where the directional components in discrete form are therefore, N1 P2 1 P2 ˜ P1 P2 ), τ1 = τN − [M]A(IPN11PP22 Q P1 P2 = [M]S

(18)

and that these directional components decrease exponentially with the polynomial order in smooth solutions. Consequently, it is possible to use a semi-converged solution q˜ P1 P2 to estimate τ N1 N2 (Ni < Pi ) and then extrapolate the directional components τi to obtain the values of τ N1 N2 for Ni > Pi . Figure 1a shows a graphical representation of the truncation error τ N1 N2 as estimated with a semiconverged solution of order P1 = P2 = 5.

2.3.2 The p-Adaptation Multigrid Scheme It has been shown that the use of FAS p-multigrid methods speeds up the computation of steady-state and unsteady solutions of the compressible NavierStokes equations [2, 8]. In addition, Rueda-Ramírez et al. [18] showed that the truncation error of an anisotropic representation can be inexpensively obtained inside an anisotropic p-multigrid cycle that performs the coarsening in one coordinate direction at a time. In fact, the second term of Eq. (18) is naturally computed in an anisotropic multigrid for obtaining the coarse-grid source term (Eq. (9)). Therefore, we propose a p-adaptation multigrid scheme that makes use of the multigrid as a solver, but also as an error estimator. Every time the error is estimated, an anisotropic p-multigrid strategy is used to generate a truncation error map for each element, like the one in Fig. 1a. Afterwards, the polynomial orders in the different coordinate directions are selected for

60

-3

50

-3.5 -4

40

-4.5 30

-5 -5.5

20

-6

10

-6.5

(a)

(b)

N1 N2 Fig. 1 (a) Truncation error map for a specific element that shows log τ5,5

∞

as a function of

(b) N1 and N2 (the black box shows the limit between the estimated and extrapolated maps). Map N1 N2 of degrees of freedom (the black boxes show the polynomial orders that achieve τ5,5 < 10−5 )

∞

An Anisotropic p-Adaptation Multigrid Scheme for the DGSEM

555

each element, such that a truncation error threshold τmax is achieved with the minimum NDOF possible, as illustrated in Fig. 1b. In the simulations shown in this paper, the reference representation, q˜ P , is converged to a residual τmax /10 before the p-adaptation stage, so that the truncation error is accurately estimated down to τmax , as was shown necessary by Kompenhans et al. [10].

3 Flow Past a NACA0012 Airfoil In this section, we compare the performance of the proposed p-adaptation multigrid scheme with a uniformly adapted p-multigrid method (without local padaptation) and a uniformly adapted RK3 method when solving the steady viscous flow past a NACA0012 airfoil at angle of attack 5◦ , Re∞ = 200 (L∞ = Lchord) and M∞ = 0.2. This particular settings correspond to a steady laminar flow, but the proposed method can be directly used with any steady solution (e.g. RANS). An unstructured mesh of 2011 quadrilateral elements is employed (Fig. 2). In the cases where multigrid is employed, the RK3 scheme is used as the iterative method (smoother), so that additional speed-ups are only due to the methods exposed in Sect. 2. As in [18], a residual-based smoothing strategy is performed. The minimum number of smoothing sweeps is β = 200 for the coarsest multigrid level (N = 1) and β = 50 for any other level. After every β presmoothing sweeps, the residual in the next (coarser) representation is checked. If N−1 N < 1.2 R R , the pre-smoothing is stopped; otherwise, β additional ∞

∞

p 18.46 18.36 18.26 18.16 18.06 17.96 17.86 17.76 17.66

-1

-0.5

0

0.5

1 x

1.5

2

2.5

3

Fig. 2 Pressure contours of the flow past a NACA0012 at angle of attack 5◦

556

A. M. Rueda-Ramírez et al.

sweeps are performed. Similarly, the norm of the residual after the post-smoothing ≤ is forced to be at least as low as it was after the pre-smoothing, RN post ∞ N Rpre . If that condition is not fulfilled, additional β sweeps are taken until it ∞ is. The isolated truncation error estimate is used to drive the p-adaptation method since it has been shown to provide better results than the non-isolated one [17– 19]. The conservative form (Eq. (1)) of the compressible Navier-Stokes equations is discretized using the Discontinuous Galerkin Spectral Element Method (DGSEM) [3, 12], which is a nodal (collocation) version of a DG method that uses Gauss points as the solution nodes and quadrature points, obtaining diagonal mass matrices. However, the methods that are exposed here can be applied to any DG scheme with tensor-product basis functions. In [18] it was explained that, when using the DGSEM in general 3D curved meshes and p-nonconforming representations, the order of the mapping must be at most M ≤ N/2 for the numerical representation to be free-stream preserving. For this reason, the use of a conforming algorithm was proposed, which forces the polynomial orders to be conforming in the first layer of elements on a curved boundary. The use of a conforming algorithm is necessary to retain the well-known M ≤ N condition of the DGSEM [11]. In this work, we use the conforming algorithm on the airfoil surface since it showed to produce better results, although its use is not imperative as the considered test case is 2D. For the uniformly adapted cases, the polynomial order is varied between N = 2 and N = 7. For the cases with local p-adaptation, a single-stage anisotropic p-adaptation procedure is performed, and the minimum polynomial order after adaptation is set to Nmin = 1, whereas the maximum polynomial order after adaptation is set to Nmax = 7. The relative drag and lift errors of the adapted meshes are assessed by comparing with a reference solution of order N = 8: N=8 edrag =

|Cd − CdN=8 | CdN=8

N=8 , elift =

|Cd − ClN=8 | ClN=8

.

(19)

Figure 3 shows a comparison between the errors obtained using the τˆ -based adaptation procedure and the ones using uniform p-refinement. As can be observed, the number of degrees of freedom is substantially reduced for the same accuracy when using the τˆ -based p-adaptation. This reduction translates into a reduction of the CPU-times. It is interesting to point out that, as the isolated truncation error threshold τˆmax is decreased, the polynomial orders of the mesh tend to the maximum specified polynomial order, Nmax = 7. Consequently, the lift and drag coefficients also tends to ClN=7 . Using Fig. 3, it is possible to compute a speed-up for different levels of accuracy. Table 1 summarizes the speed-up calculations for the maximum level of accuracy that was achieved for the drag and lift coefficients.

An Anisotropic p-Adaptation Multigrid Scheme for the DGSEM

557

(a)

(b)

(c)

(d)

Fig. 3 Relative error in the drag and lift coefficients for different methods for the flow past the NACA0012 airfoil. The blue lines represent uniform refinement, and the red lines represent the τˆ -based p-adaptation procedure with Nmax = 7. (a) Drag error vs. DOFs; (b) lift error vs. DOFs; (c) drag error vs. CPU-time; (d) lift error vs. CPU-time Table 1 Computation times and speed-up for the different methods after converging until r∞ < 10−9 Method RK3 FAS FAS + padaptation

Drag coefficient (edrag ≤ ×4.1 × 10−5 ) CPU-time [s] Time [%] Speed-up 1.95 × 107 100.00% 1.00 2.36 × 106 12.10% 8.26 1.21 × 106 6.20% 16.13

Lift coefficient (elift ≤ 2.4 × 10−5 ) CPU-time [s] Time [%] Speed-up 1.95×107 100.00% 1.00 2.36×106 12.10% 8.26 1.48×106 7.58% 13.19

Figure 4 shows the distribution of polynomial orders after the single-stage adaptation procedure for a threshold of τmax = 5 × 10−4 , which has related errors N=8 N=8 of edrag = 4.10 × 10−5 and elift = 7.31 × 10−5 . As can be observed, the elements that are enriched are mainly the ones on the boundary layer (specially leading and trailing edge), and the zones of the wake where the element size changes significantly.

558

A. M. Rueda-Ramírez et al.

Naverage 6.5 6 5.5 5 4.5 4 3.5 3 2.5 2 1.5

Fig. 4 Polynomial order distribution after the anisotropic p-adaptation. Naverage = (N1 + N2 )/2

4 Conclusions In this work, we have applied recently developed error estimators and anisotropic p-adaptation methods in conjunction with multigrid solving strategies for solving the compressible Navier-Stokes equations. In particular, we have shown that the coupling of anisotropic truncation error-based p-adaptation methods with pmultigrid schemes can speed up the computation of steady-state solutions of PDEs. The achieved speed-up depends on the desired accuracy, being this method optimal when high accuracy is required (low errors). In particular, a speed-up of 16.13 was achieved for the computation of the steady compressible viscous flow past a NACA0012 airfoil at angle of attack 5◦ with respect to the uniformly adapted representation without multigrid. Acknowledgements This project has received funding from the European Union’s Horizon 2020 Research and Innovation Program under the Marie Skłodowska-Curie grant agreement No 675008 for the SSeMID project. The authors acknowledge the computer resources and technical assistance provided by the Centro de Supercomputación y Visualización de Madrid (CeSViMa).

References 1. Bassi, F., Rebay, S.: A high-order accurate discontinuous finite element method for the numerical solution of the compressible Navier-Stokes equations. J. Comput. Phys. 131, 267– 279 (1997) 2. Bassi, F., Ghidoni, A., Rebay, S., Tesini, P.F.: High-order accurate p-multigrid discontinuous Galerkin solution of the Euler equations. Int. J. Numer. Methods Fluids 60, 847–865 (2009) 3. Black, K.: A conservative spectral element method for the approximation of compressible fluid flow. Kybernetika 35, 133–146 (1999)

An Anisotropic p-Adaptation Multigrid Scheme for the DGSEM

559

4. Brandt, A., Livne, O.E.: Multigrid Techniques: 1984 Guide with Applications to Fluid Dynamics, Revised Edition. SIAM, Philadelphia (2011) 5. Cockburn, B., Shu, C.-W.: The local discontinuous Galerkin method for time-dependent convection-diffusion systems. SIAM J. Numer. Anal. 35, 2440–2463 (1998) 6. Ferrer, E.: An interior penalty stabilised incompressible discontinuous Galerkin–Fourier solver for implicit large eddy simulations. J. Comput. Phys. 348, 754–775 (2017) 7. Ferrer, E., Willden, R.H.: A high order Discontinuous Galerkin–Fourier incompressible 3D Navier-Stokes solver with rotating sliding meshes. J. Comput. Phys. 231, 7037–7056 (2012) 8. Fidkowski, K.J., Oliver, T.A., Lu, J., Darmofal, D.L.: p-Multigrid solution of high-order discontinuous Galerkin discretizations of the compressible Navier-Stokes equations. J. Comput. Phys. 207, 92–113 (2005) 9. Fraysse, F., Rubio, G., De Vicente, J., Valero, E.: Quasi-a priori mesh adaptation and extrapolation to higher order using τ -estimation. Aerosp. Sci. Technol. 38, 76–87 (2014) 10. Kompenhans, M., Rubio, G., Ferrer, E., Valero, E.: Adaptation strategies for high order discontinuous Galerkin methods based on Tau-estimation. J. Comput. Phys. 306, 216–236 (2016) 11. Kopriva, D.A.: Metric identities and the discontinuous spectral element method on curvilinear meshes. J. Sci. Comput. 26, 301–327 (2006) 12. Kopriva, D.: Implementing Spectral Methods for Partial Differential Equations: Algorithms for Scientists and Engineers. Springer, Berlin (2009) 13. Kopriva, D.A., Woodruff, S.L., Hussaini, M.Y.: Computation of electromagnetic scattering with a non-conforming discontinuous spectral element method. Int. J. Numer. Methods Eng. 53, 105–122 (2002) 14. Mitchell, W.F., Division, C.S.: The hp -multigrid method applied to hp -adaptive refinement of triangular grids. Numer. Linear Algebra Appl. 17, 211–228 (2010) 15. Rivière, B.: Discontinuous Galerkin Methods for Solving Elliptic and Parabolic Equations Theory and Implementation. SIAM, Philadelphia (2008) 16. Roe, P.L.: Approximate Riemann solvers, parameter vectors, and difference schemes. J. Comput. Phys. 43, 357–372 (1981) 17. Rubio, G., Fraysse, F., Kopriva, D.A., Valero, E.: Quasi-a priori truncation error estimation in the DGSEM. J. Sci. Comput. 64, 425–455 (2015) 18. Rueda-Ramírez, A.M., Manzanero, J., Ferrer, E., Rubio, G., Valero, E.: A p-multigrid strategy with anisotropic p-adaptation based on truncation errors for high-order discontinuous Galerkin methods. J. Comput. Phys. 378, 209–233 (2019) 19. Rueda-Ramírez, A.M., Rubio, G., Ferrer, E., Valero, E.: Truncation error estimation in the panisotropic discontinuous Galerkin spectral element method. J. Sci. Comput. 78(1), 433–466 (2018) 20. Syrakos, A., Efthimiou, G., Bartzis, J.G., Goulas, A.: Numerical experiments on the efficiency of local grid refinement based on truncation error estimates. J. Comput. Phys. 231, 6725–6753 (2012) 21. Wang, L., Mavriplis, D.: Adjoint-based h-p adaptive discontinuous Galerkin methods for the compressible Euler equations, J. Comput. Phys. 228, 7643–7661 (2009) 22. Wang, Z., Fidkowski, K., Abgrall, R., Bassi, F., Caraeni, D., Cary, A., Deconinck, H., Hartmann, R., Hillewaert, K., Huynh, H., Kroll, N., May, G., Persson, P.-O., van Leer, B., Visbal, M.: High-order CFD methods: current status and perspective. Int. J. Numer. Methods Fluids 72, 811–845 (2013)

560

A. M. Rueda-Ramírez et al.

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

A Spectral Element Reduced Basis Method for Navier–Stokes Equations with Geometric Variations Martin W. Hess, Annalisa Quaini, and Gianluigi Rozza

1 Introduction and Motivation Spectral element methods (SEM) use high-order polynomial ansatz functions to solve partial differential equations (PDEs) in all fields of science and engineering, see, e.g., [4–7, 12, 16] and references therein for an overview. Typically, an exponential error decay under p-refinement is observed, which can provide an enhanced accuracy over standard finite element methods at the same computational cost. In the following, we assume that the discretization error is much smaller than the model reduction error, small enough not to interfere with our results. In general, this needs to be established with the use of suitable error estimation and adaptivity techniques. We consider the flow through a channel with a narrowing of variable height. A reduced order model (ROM) is computed from a few high-order SEM solves, which accurately approximates the high-order solutions for the parameter range of interest, i.e., the different narrowing heights under consideration. Since the parametric variations are affine, a mapping to a reference domain is applied without further interpolation techniques. The focus of this work is to show how to use simulations arising from the SEM solver Nektar++ [3] in a ROM context. In particular, the multilevel static condensation of the high-order solver is not applied, but the ROM projection works with the system matrices in local coordinates. See [12] for further details. This is in contrast to our previous work [8], since numerical

M. W. Hess () · G. Rozza SISSA mathLab, International School for Advanced Studies, Trieste, Italy e-mail: [email protected]; [email protected] A. Quaini Department of Mathematics, University of Houston, Houston, TX, USA e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_45

561

562

M. W. Hess et al.

experiments have shown that the multilevel static condensation is inefficient in a ROM context. Additionally, we consider affine geometry variations. With SEM as discretization method, we use global approximation functions for the high-order as well as reduced-order methods. The ROM techniques described in this paper are implemented in open-source project ITHACA-SEM.1 The outline of the paper is as follows. In Sect. 2, the model problem is defined and the geometric variations are introduced. Section 3 provides details on the spectral element discretization, while Sect. 4 describes the model reduction approach and shows the affine mapping to the reference domain. Numerical results are given in Sect. 5, while Sect. 6 summarizes the work and points out future perspectives.

2 Problem Formulation Let  ∈ R2 be the computational domain. Incompressible, viscous fluid motion in spatial domain  over a time interval (0, T ) is governed by the incompressible Navier-Stokes equations with vector-valued velocity u, scalar-valued pressure p, kinematic viscosity ν and a body forcing f: ∂u + u · ∇u = −∇p + ν u + f, ∂t ∇ · u = 0.

(1) (2)

Boundary and initial conditions are prescribed as u=d

on D × (0, T ),

(3)

∇u · n = g

on N × (0, T ),

(4)

u = u0

in  × 0,

(5)

with d, g and u0 given and ∂ = D ∪ N , D ∩ N = ∅. The Reynolds number Re, which characterizes the flow [11], depends on ν, a characteristic velocity U , and a characteristic length L: Re =

UL . ν

(6)

We are interested in computing the steady states, i.e., solutions where ∂u ∂t vanishes. The high-order simulations are obtained through time-advancement, while the ROM solutions are obtained with a fixed-point iteration.

1 https://github.com/mathLab/ITHACA-SEM.

SEM-RB for NSE with Geometric Variations

563

2.1 Oseen-Iteration The Oseen-iteration is a secant modulus fixed-point iteration, which in general exhibits a linear rate of convergence [2]. Given a current iterate (or initial condition) uk , the next iterate uk+1 is found by solving linear system: −ν uk+1 + (uk · ∇)uk+1 + ∇p = f in , ∇ · uk+1 = 0 in , uk+1 = d

on D ,

∇uk+1 · n = g

on N .

Iterations are typical stopped when the relative difference between iterates falls below a predefined tolerance in a suitable norm, like the L2 () or H01 () norm.

2.2 Model Description We consider the reference computational domain shown in Fig. 1, which is decomposed into 36 triangular spectral elements. The spectral element expansion uses modal Legendre polynomials of the Koornwinder-Dubiner type of order p = 11 for the velocity. Details on the discretization method can be found in chapter 3.2 of [12]. The pressure ansatz space is chosen of order p − 2 to fulfill the inf-sup stability condition [1, 20]. A parabolic inflow profile is prescribed at the inlet (i.e., x = 0) with horizontal velocity component ux (0, y) = y(3 − y) for y ∈ [0, 3]. At the outlet (i.e., x = 8) we impose a stress-free boundary condition, everywhere else we prescribe a no-slip condition. The height of the narrowing in the reference configuration is μ = 1, from y = 1 to y = 2. See Fig. 1. Parameter μ is considered variable in the interval μ ∈ [0.1, 2.9]. The narrowing is shrunken or expanded as to maintain the geometry

Fig. 1 Reference computational domain for the channel flow, divided into 36 triangles

564

M. W. Hess et al.

Fig. 2 Full order, steady-state solution for μ = 1: velocity in x-direction (top) and y-direction (bottom)

Fig. 3 Full order, steady-state solution for μ = 0.1: velocity in x-direction (top) and y-direction (bottom)

symmetric about line y = 1.5. Figures 2, 3, and 4 show the velocity components close to the steady state for μ = 1, 0.1, 2.9, respectively. The viscosity is kept constant to ν = 1. For these simulations, the Reynolds number (6) is between 5 and 10, with maximum velocity in the narrowing as characteristic velocity U and the height of the narrowing characteristic length L. For larger Reynolds numbers (about 30), a supercritical pitchfork bifurcation occurs giving rise to the so-called Coanda effect [8, 9, 22], which is not subject of the current study. Our model is similar to the model considered in [17, 18], i.e. an

SEM-RB for NSE with Geometric Variations

565

Fig. 4 Full order, steady-state solution for μ = 2.9: velocity in x-direction (top) and y-direction (bottom)

expansion channel with an inflow profile of varying height. However, in [18] the computational domain itself does not change.

3 Spectral Element Full Order Discretization The Navier-Stokes problem is discretized with the spectral element method. The spectral/hp element software framework used is Nektar++ in version 4.4.0.2 The discretized system of size Nδ to solve at each step of the Oseen-iteration for fixed μ can be written as ⎡ ⎤ ⎡ ⎤ ⎤⎡ T A −Dbnd fbnd B vbnd ⎢ ⎥ ⎢ ⎥ ⎥⎢ (7) 0 −Dint ⎦ ⎣ p ⎦ = ⎣ 0 ⎦ , ⎣ −Dbnd T C B˜ T −Dint vint fint where vbnd and vint denote velocity degrees of freedom on the boundary and in the interior of the domain, respectively, while p denotes the pressure degrees of freedom. The forcing terms on the boundary and interior are denoted by fbnd and fint , respectively. The matrix A assembles the boundary-boundary coupling, B the boundary-interior coupling, B˜ the interior-boundary coupling, and C assembles the interior-interior coupling of elemental velocity ansatz functions. In the case of a Stokes system, it holds that B = B˜ T , but this is not the case for the Oseen equation because of the linearized convective term. The matrices Dbnd

2 See

www.nektar.info.

566

M. W. Hess et al.

and Dint assemble the pressure-velocity boundary and pressure-velocity interior contributions, respectively. The linear system (7) is assembled in local degrees of freedom, resulting in ˜ C, Dbnd and Dint , each block corresponding to a spectral block matrices A, B, B, element. This allows for an efficient matrix assembly since each spectral element is independent from the others, but makes the system singular. In order to solve the system, the local degrees of freedom need to be gathered into the global degrees of freedom [12]. The high-order element solver Nektar++ uses a multilevel static condensation for the solution of linear systems like (7). Since static condensation introduces intermediate parameter-dependent matrix inversions (such as C −1 in this case) several intermediate projection spaces need to be introduced to use model order reduction [8]. This can be avoided by instead projecting the expanded system (7) directly. The internal degrees of freedom do not need to be gathered, since they are the same in local and global coordinates. Only ansatz functions extending over multiple spectral elements need to be gathered. Next, we will take the boundary-boundary coupling across element interfaces into account. Let M denote the rectangular matrix which gathers the local boundary degrees of freedom into global boundary degrees of freedom. Multiplication of the first row of (7) by M T M will then set the boundary-boundary coupling in local degrees of freedom: ⎤ ⎡ ⎤ ⎤⎡ T M T Mfbnd M T MA −M T MDbnd vbnd M T MB ⎢ ⎥ ⎢ ⎥ ⎥⎢ 0 −Dint ⎦ ⎣ p ⎦ = ⎣ 0 ⎣ −Dbnd ⎦. T T ˜ vint fint B −Dint C ⎡

(8)

The action of the matrix in (8) on the degrees of freedom on the Dirichlet boundary is computed and added to the right hand side. Such degrees of freedom are then removed from (8). The resulting system can then be used in a projectionbased ROM context [13], of high-order dimension Nδ × Nδ and depending on the parameter μ: A(μ)x(μ) = f.

(9)

4 Reduced Order Model The reduced order model (ROM) computes accurate approximations to the highorder solutions in the parameter range of interest, while greatly reducing the overall computational time. This is achieved by two ingredients. First, a few high-order solutions are computed and the most significant proper orthogonal decomposition (POD) modes are obtained [13]. These POD modes define the reduced order ansatz space of dimension N, in which the system is solved. Second, to reduce

SEM-RB for NSE with Geometric Variations

567

the computational time, an offline-online computational procedure is used. See Sect. 4.1. The POD computes a singular value decomposition of the snapshot solutions to 99.99% of the most dominant modes [10], which define the projection matrix U ∈ RNδ ×N used to project system (9): U T A(μ)U xN (μ) = U T f.

(10)

The low order solution xN (μ) then approximates the high order solution as x(μ) ≈ U xN (μ).

4.1 Offline-Online Decomposition The offline-online decomposition [10] enables the computational speed-up of the ROM approach in many-query scenarios. It relies on an affine parameter dependency, such that all computations depending on the high-order model size can be moved into a parameter-independent offline phase, while having a fast inputoutput evaluation online. In the example under consideration here, the parameter dependency is already affine and a mapping to the reference domain can be established without using an approximation technique such as the empirical interpolation method. Thus, there exists an affine expansion of the system matrix A(μ) in the parameter μ as A(μ) =

Q 

:i (μ)Ai .

(11)

i=1

The coefficients :i (μ) are computed from the mapping x = Tk (μ)ˆx + gk , ˆ k to the reference Tk ∈ R2×2 , gk ∈ R2 , which maps the deformed subdomain  subdomain k . See also [19, 21]. Figure 5 shows the reference subdomains k for the problem under consideration. ˆ k the elemental basis function evaluations are transformed For each subdomain  to the reference domain. For each velocity basis function u = (u1 , u2 ), v = (v1 , v2 ), w = (w1 , w2 ) and each (scalar) pressure basis function ψ, we can write the transformation with summation convention as:   ∂ uˆ ∂ vˆ ∂u ∂v ˆ νˆ ij d k = νij dk , ∂ x ˆ ∂ x ˆ ∂x ∂x ˆk i j i j k    ∂uj ˆ · ud ˆk = ˆ  ψ∇ ψχij dk , ∂xi ˆ k k   ∂vj ˆk = ˆ  (uˆ · ∇)ˆv · wd ui πij wdk , ∂xi ˆk k 

568

M. W. Hess et al.

Fig. 5 Reference computational domain with subdomains 1 (green), 2 (yellow), 3 (blue), 4 (grey) and 5 (brown)

with νij = Tii  νˆi  j  Tjj  det(T )−1 , χij = πij = Tij det(T )−1 . The subdomain 5 (see Fig. 5) is kept constant, so that no interpolation of the inflow profile is necessary. To achieve fast reduced order solves, the offline-online decomposition expands the system matrix as in (11) and computes the parameter independent projections offline, which are stored as small-sized matrices of the order N × N. Since in an Oseen-iteration each matrix is dependent on the previous iterate, the submatrices corresponding to each basis function are assembled and then formed online using the reduced basis coordinate representation of the current iterate. This is the same procedure used for the assembly of the nonlinear term in the Navier-Stokes case [13].

5 Numerical Results The accuracy of the ROM is assessed using 40 snapshots sampled uniformly over the parameter domain [0.1, 2.9] for the POD and 40 randomly chosen parameter locations to test the accuracy. Figure 6 (left) shows the decay of the energy of the POD modes. To reach the typical threshold of 99.99% on the POD energy, it takes 9 POD modes as RB ansatz functions. Figure 6 (right) shows the relative L2 () approximation error of the reduced order model with respect to the full order model up to 6 digits of accuracy, evaluated at the 40 randomly chosen verification parameter locations. With 9 POD modes the maximum approximation error is less than 0.7% and the mean approximation error is less than 0.5%. While the full-order solves were computed with Nektar++, the reduced-order computations were done in ITHACA-SEM with a separate python code. To assess the computational gain, the time for a fixed point iteration step using the full-

104 100 10−4 10−80

10

20 30 RB dimension N

40

569 Relative L2 () velocity error

POD mode energy

SEM-RB for NSE with Geometric Variations

100 10−2 10−4 10−6 0

5

10 15 RB dimension N

20

Fig. 6 Left: Decay of POD mode energy. Right: Maximum (red) and mean (blue) relative L2 () error for the velocity over increasing reduced basis dimension

order system is compared to the time for a fixed point iteration step of the ROM with dimension 20, both done in python. The ROM online phase reduces the computational time by a factor of over 100. The offline time is dominated by computing the snapshots and the trilinear forms used to project the advection terms. See [13] for detailed explanations.

6 Conclusion and Outlook We showed that the POD reduced basis technique generates accurate reduced order models for SEM discretized models under parametric variation of the geometry. The potential of a high-order spectral element method with a reduced basis ROM is the subject of current investigations. See also [6]. Since each spectral element comprises a block in the system matrix in local coordinates, a variant of the reduced basis element method (RBEM) [14, 15] can be successfully applied in the future. Acknowledgements This work was supported by European Union Funding for Research and Innovation through the European Research Council (project H2020 ERC CoG 2015 AROMACFD project 681447, P.I. Prof. G. Rozza). This work was also partially supported by NSF through grant DMS-1620384 (Prof. A. Quaini).

References 1. Boffi, D., Brezzi F., Fortin, M.: Mixed Finite Element Methods and Applications. Springer Series in Computational Mathematics. Springer, Berlin (2013) 2. Burger, M.: Numerical Methods for Incompressible Flow. Lecture Notes. UCLA, Los Angeles (2010)

570

M. W. Hess et al.

3. Cantwell, C.D., Moxey, D., Comerford, A., Bolis, A., Rocco, G., Mengaldo, G., de Grazia, D., Yakovlev, S., Lombard, J.-E., Ekelschot, D., Jordi, B., Xu, H., Mohamied, Y., Eskilsson, C., Nelson, B., Vos, P., Biotto, C., Kirby, R.M., Sherwin, S.J.: Nektar++: an open-source spectral/hp element framework. Comput. Phys. Commun. 192, 205–219 (2015) 4. Canuto, C., Hussaini, M.Y., Quarteroni, A., Zhang, Th.A.: Spectral Methods Fundamentals in Single Domains. Scientific Computation. Springer, Berlin (2006) 5. Canuto, C., Hussaini, M.Y., Quarteroni, A., Zhang, Th.A.: Spectral Methods Evolution to Complex Geometries and Applications to Fluid Dynamics. Scientific Computation. Springer, Berlin (2007) 6. Fick, L., Maday, Y., Patera A., Taddei T.: A stabilized POD model for turbulent flows over a range of Reynolds numbers: optimal parameter sampling and constrained projection. J. Comput. Phys. 371, 214–243 (2018) 7. Herrero, H., Maday, Y., Pla, F.: RB (Reduced Basis) for RB (Rayleigh–Bénard). Comput. Meth. Appl. Mech. Eng. 261–262, 132–141 (2013) 8. Hess, M.W., Rozza, G.: A spectral element reduced basis method in parametric CFD. In: Numerical Mathematics and Advanced Applications ENUMATH 2017. Springer, Berlin (2018, in press). E-print arXiv:1712.06432 9. Hess, M.W., Alla, A., Quaini, A., Rozza, G., Gunzburger, M.: A localized reducedorder modeling approach for PDEs with bifurcating solutions. In: Computer Methods in Applied Mechanics and Engineering (CMAME) (2019, accepted for publication). E-print. arXiv:1807.08851 10. Hesthaven, J.S., Rozza, G., Stamm, B.: Certified Reduced Basis Methods for Parametrized Partial Differential Equations. SpringerBriefs in Mathematics. Springer, Berlin (2016) 11. Holmes, P., Lumley, J., Berkooz, G.: Turbulence, Coherent Structures, Dynamical Systems and Symmetry. Cambridge University Press, Cambridge (1996) 12. Karniadakis, G., Sherwin, S.: Spectral/hp Element Methods for Computational Fluid Dynamics, 2nd edn. Oxford University Press, Oxford (2005) 13. Lassila, T., Manzoni, A., Quarteroni, A., Rozza, G.: Model order reduction in fluid dynamics: Challenges and perspectives. In: Quarteroni, A., Rozza, G. (eds.) Reduced Order Methods for Modelling and Computational Reduction. MS&A Modeling, Simulation and Applications, vol. 9, pp. 235–273. Springer International Publishing, Cham (2014) 14. Lovgren, A.E., Maday, Y, Ronquist, E.M.: A reduced basis element method for the steady Stokes problem. ESAIM: Math. Model. Numer. Anal. 40(3), 529–552 (2006) 15. Maday, Y., Ronquist, E.M.: A reduced-basis element method. Comptes Rendus Math. 335(2), 195–200 (2002) 16. Patera, A.T.: A spectral element method for fluid dynamics; laminar flow in a channel expansion. J. Comput. Phys. 54(3), 468–488 (1984) 17. Pitton, G., Rozza, G.: On the application of reduced basis methods to bifurcation problems in incompressible fluid dynamics. J. Sci. Comput. 73, 157–177 (2017) 18. Pitton, G., Quaini, A., Rozza, G.: Computational reduction strategies for the detection of steady bifurcations in incompressible fluid-dynamics: applications to Coanda effect in cardiology. J. Comput. Phys. 344, 534–557 (2017) 19. Quarteroni, A., Rozza, G.: Numerical solution of parametrized Navier-Stokes equations by reduced basis methods. Num. Meth. Part. Diff. Eq. 23(4), 923–948 (2007) 20. Quarteroni, A., Valli, A.: Numerical Approximation of Partial Differential Equations. Springer, Berlin (1994) 21. Rozza, G.: Real-time reduced basis solutions for Navier-Stokes equations: optimization of parametrized bypass configurations. In: ECCOMAS CFD 2006 Proceedings on CD, vol. 676, pp. 1–16 (2006) 22. Wille, R., Fernholz, H.: Report on the first European mechanics colloquium, on the Coanda effect. J. Fluid Mech. 23(4), 801–819 (1965)

SEM-RB for NSE with Geometric Variations

571

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Iterative Spectral Mollification and Conjugation for Successive Edge Detection Robert E. Tuzun and Jae-Hun Jung

1 Introduction Detection of edges is a fundamental problem in a variety of applications, including image processing and the numerical solution of differential equations. In applications such as magnetic resonance imaging (MRI), it is required to construct images from Fourier data. Let {fˆk |k = 0, ±1, ±2, · · · } be the set of Fourier coefficients of f (x) ∈ L2 [−π, π] given by 1 fˆk = 2π



π −π

f (x)e−ikx dx,

< ˆ ikx . When the underlying and let fN be the Fourier partial sum fN = N k=−N fk e function is smooth and periodic, the Fourier reconstruction fN is accurate to spectral accuracy, but when edges are present, the reconstruction is plagued by the Gibbs phenomenon, also known as the Gibbs ringing in MRI applications. Various methods have been proposed to address these issues and those methods consist of edge detection followed by reconstruction. Thus, the determination of edge locations is critical. Fourier concentration method has emerged over the past

R. E. Tuzun Department of Mathematics, University at Buffalo, The State University of New York, Buffalo, NY, USA e-mail: [email protected] J.-H. Jung () Department of Mathematics, University at Buffalo, The State University of New York, Buffalo, NY, USA Department of Data Science, Ajou University, Suwon, South Korea e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_46

573

574

R. E. Tuzun and J.-H. Jung

decade as a robust method for edge detection in a variety of circumstances and applications [5, 6]. Essentially, a certain Fourier partial sum converges to the jump function as the number of Fourier coefficients increases and this convergence can be accelerated by what is known as concentration factors (functions). Use of different types of concentration factors tends to impart trade-offs between oscillations near jump discontinuities and significant non-zero concentration away from them [2]. Several methods have been devised to address this issue, as well as to treat special circumstances such as incomplete Fourier data and the presence of noise [1, 4, 13, 15]. Thanks to the convergence property of the Fourier concentration to the jump function, the concentration method detects edges with large concentrations. Where the function is smooth, the concentration vanishes as the jump function vanishes as N → ∞. In practice, the concentration method is designed to detect edges with magnitudes larger than some given threshold, with the value of the used threshold being problem dependent. The value of the threshold cannot be arbitrarily small; otherwise, too many false edges can be detected. If the magnitude of weak edges is much smaller than other edges, those edges are considered insignificant, but for some cases weak edges are more important than strong edges. For example, it was shown that in the segmentation of MRI of the knee, the cartilage is better characterized by weak edges rather than strong edges for the separation from the tibia and femur [11]. This note shows that an iterative approach based on the successive conjugation and adaptive mollification can detect all edges without any prior threshold. This approach is similar to the iterative method in the context of the radial basis function method [3, 9, 10]. The iterative method is as follows: at each iteration step, all previously found edges are smoothed by a local mollification and new corresponding Fourier coefficients are computed. By applying conjugation and mollification successively, one can distinguish real edges from fake edges. This approach is useful and effective particularly for problems where the weak jump can significantly affect the global solution of differential equations or images where the interesting structure is represented by the weak edges [11]. In Sect. 2, a brief explanation of the Fourier concentration method is given. In Sect. 3, the proposed iterative method is explained based on the adaptive filtering method. The stopping criteria is also explained. Numerical examples with remarks are given. In Sect. 4, a brief concluding remark is provided.

2 Edge Detection Using Fourier Concentration Method Let [f ](x) = f (x + ) − f (x − ) denote the jump function of f (x) ∈ L2 [−π, π], where the superscripts + and − denote the limits taken from the right and left, respectively. Given a finite set of Fourier coefficients, {fˆk }|k|≤N , the Fourier

Iterative Spectral Mollification and Conjugation for Successive Edge Detection

575

concentration method, developed in [5, 6], computes the concentration as a sum of the form   |k| ikx σ SN [f ](x) = i sgn(k)fˆk σ (1) e N |k|≤N

where the σ (·) are known as concentration factors and sgn(k) is the sign function. Given certain admissibility conditions [6], the sum converges to the jump function: ⎧   ⎪ ⎨ O log N , d(x) ≤ logNN N σ   SN [f ](x) = [f ](x) + (2) log N ⎪ ⎩ O (Nd(x))s , d(x) ' N −1 where d(x) denotes the distance to the nearest edge and s depends on the concentration factor. Here we note that Eq. (2) shows that the concentration function σ [f ](x) recovers the jump function of f (x) as N → ∞ and the convergence SN may be slow. Equation (2) also implies that the absolute maximum value of the σ [f ](x) converges to the maximum jump. Accordingly concentration function SN we observe that strong jumps are relatively easier to detect than weak jumps. The common types of concentration factors satisfying the admissibility conditions are polynomial concentrations σ (η) = pηp ,

p ≥ 1,

(3)

where p is a positive integer and η = |k|/N and exponential concentration functions σ (η) = Cηe1/(αη(η−1)),

(4)

where α > 0 is an order and C is a normalization constant. Cutoffs for edge detection, τ ∈ (0, 1], are with respect to the normalized concentration σ σ ˆ [f ](y)|/ max{|SN [f ](y)|} S(y) = |SN y

and the edge set, E, is defined as ˆ E = {x|S(x) ≥ τ, x ∈ [−π, π]}.

(5)

Several approaches have been developed for improving the concentration method. We refer readers some to [1, 2, 4, 6, 7, 12, 13, 15, 16]. All these methods are basically utilizing the edge map. Figure 1 is by the Fourier concentration method for f1 (x) ⎧ ⎪ 3 π4 ≤ x ≤ π2 ⎪ ⎪ ⎨ −2 π ≤ x ≤ 54 π f1 (x) = ⎪ mw 32 π ≤ x ≤ 74 π ⎪ ⎪ ⎩ 0 otherwise

(6)

R. E. Tuzun and J.-H. Jung 1

3

0.9

2.5

0.8

2

0.7

1.5

0.6

1

f(x)

Concentration Ratio

576

0.5 0.4

0

0.3

−0.5

0.2

−1

0.1

−1.5

0

−2 0

1

2

4

3

5

6

0

7

x

a) 1

3 2.5

0.8

2

0.7

1.5

0.6

3

4

5

6

4

5

6

1

0.5

0.5

0.4

0

0.3

−0.5

0.2

−1

0.1

−1.5

c)

2

x

0.9

0

1

b)

f(x)

Concentration Ratio

0.5

−2 0

1

2

3

4

x

5

6

7

0

d)

1

2

3

x

Fig. 1 Edge detection for f1 (x) with τ = 0.4. (a and b) with the polynomial concentration9with p = 5. (c :and d) with the exponential concentration with p = 12 for σ =    p   ˆ (b and d) the detected edges marked and #M = 64. (a and c) S(y). exp −#M 1 −  Nk  by red cross symbols. N = 128. The red cross in (b and d) indicates the edge locations found

where mw = 0.1. That is, the magnitude of the strong jump is 30 times the weakest. As clearly shown in the figure, it is hard to detect the weak edge by looking at ˆ the normalized Fourier concentration S(y), (a) and (c) in Fig. 1, although the weak edges are clearly visible in (b) and (d) in Fig. 1.

Iterative Spectral Mollification and Conjugation for Successive Edge Detection

577

3 Iterative Concentration Method As clearly seen in Fig. 1, the Fourier concentration method may fail to detect the weak edge when the concentration of the weak edge is too small compared to the strong edge. To find all edges, we propose to apply the Fourier concentration method iteratively based on the local mollification using the local adaptive filtering method.

3.1 Local Adaptive Mollification The local adaptive mollification is a key step for the iterative algorithm. Consider a smooth function φ ∈ C0∞ [0, 2π] which is compactly supported such that 1 φ# (x) = φ #

 x , #

(7)

lim#→0+ φ# (x) = δ(x). Here δ(x) is the Dirac delta function. And further ;where ∞ φ(x)dx = 1. With these properties, the limit property is given by −∞ lim (φ# ∗ f )(x) = f (0),

#→0+

where (∗) operation denotes convolution. The parameter # is free and it localizes the convolution and is known as the localization factor. The parameter # is a fixed value for every x. Thus a global smoothing occurs everywhere including both the nonsmooth and smooth areas. However, we only want to apply the mollification locally to minimize the Gibbs oscillations near the jump. In order to achieve this, we use a two-parameter family of the spectral mollifier introduced by Gottlieb and Tadmor [8]. Consider the convolution of the Fourier partial sum fN (x) and the mollifier φ. Then by the definition of fN (x) and φ we have (φ# ∗ fN (x)) =

1 2π

1 = 2π



2π

φ# (x − y)fN (y)dy

0



2π

φ# (x − y)(DN (y − z) ∗ f (z))(y)dzdy,

0

where DN is the Dirichlet kernel of degree N. The idea proposed in [8] is that one changes the degree of the Dirichlet kernel with the localization parameter # so that the two-parameter family of the new mollifier is defined by φp,# (x) =

1 ρ #

  x x Dp , # #

(8)

578

R. E. Tuzun and J.-H. Jung

where Dp is the Dirichlet kernel of degree p. Then for all s, the error is given by [8] 9 |φp,# ∗ fN (x) − f (x)| ≤ C||ρ

(s)



||∞ N

1+p N#

s+1

:  s 2 s , +p ||f ||L∞ loc p (9)

where || · ||L∞ = sup(x−#π,x+#π) | · |. The first term in the right hand side of the loc above inequality is the truncation error and the second term is the regularization error. As we see, the optimization of the error is determined by how the localization parameter # and the degree p are balanced. In [8] those parameters were chosen such that # = πd , where d is the distance to the nearest jump from the current position. The order of √ the Dirichlet kernel is chosen such that spectral convergence is achieved, say, p ≈ N. Here we note that a modification of the two-parameter mollifier for the enhancement of the convergence was proposed in [14], which was designed to reduce the Gibbs oscillations while it provides a sharp reconstruction up to the edge. Note that the adaptive mollifier was used to sharpen the concentration map Sˆ in [2]. Our proposed iterative method is that once the edge is identified, the edge region is first localized using the value of # so that fN in the region away from the detected edge is not affected by the mollification. This helps the next available edge to be preserved through the mollification of fN if existent. Thus as in [8], the localization factor is a function of the distance from the edge, d, i.e. # = #(d). Then we adaptively mollify fN so that a heavy mollification using p is applied to reduce the Gibbs oscillations near the edge. The limit property of p is given as p → 0 if d → 0 and p → ∞ if d → 2π. In this work, we use the local adaptive filtering for the mollification.

3.2 Almost Automatic Stopping of the Iteration To see the proposed method stops almost automatically, consider f (x) = x, x ∈ [−π, π] with the Fourier coefficients i fˆk = (−1)k , k

k = 0

and fˆ0 = 0. There are two edges (x = ±π) and conjugation and local adaptive filtering have the most effect at ±π. Therefore, by considering the local behavior near ±π, we assume a constant order of filtering p and of conjugation q, with functional forms of exp(−#M (1 − |k/N|)p ) and exp(−#M |k/N|q ), respectively. σ be the corresponding kernels and letting S and F denote By letting φp,# and CN conjugation and filtering, σ σ CN ∗ (φp,# ∗ fN ) ≈ φp,# ∗ (CN ∗ fN )

Iterative Spectral Mollification and Conjugation for Successive Edge Detection

579

and after some simplification, we have S[F [fN ]] =

σ CN

∗ (φp,# ∗ fN ) ≈

N  k=−N,k=0

# 

$ |k| p (−1)k+1 exp −# 1 − k N

× exp(−#|k/N|q )eikx . This has Fourier coefficients # 

p $ k+1 (−1) |k| B exp(−#|k/N|q ). exp −# 1 − S[F [fN ]] ≈ k N From the sharp localization and heavy filtering near ±π, we choose p, q −→ 0. Then setting y = |k/N| yields 1 |B S[F [fN ]]| ≈ exp[−#(y q ) + (1 − y)p ], k

k = 0

which approaches 0 exponentially. Thus after all the edges are found through iteration, the concentration decays exponentially small. Thus if the stopping criteria η below is chosen small enough, e.g. η ∼ 10−10, the stopping of the iteration is guaranteed |S[F [fN ]]| ≤ η.

(10)

3.3 Numerical Examples We consider the case that the magnitude of the weak edge is highly small for the function f1 (x) in Eq. (6) mw = 0.01. Figure 2 shows how the iteration method finds edges. The order of the finding the edges is from left to right (see red arrows). As shown in the figure, the iterative method finds all edges even with mw = 0.01. It is interesting to observe that the weakest edges are found in the 3rd and 7th iteration steps before all the strong edges are found. Now we consider even smaller value of mw mw = 0.001.

580

R. E. Tuzun and J.-H. Jung

0.6

0.6

0.6

0.6

0.6

0.6

0.4

0.4

0.4

0.4

0.4

0.4

0.2

0.2

0.2

0.2

0.2

0.2

0

0

0

0

0

0

0

2

4

6

0

2

4

6

0

2

4

6

0

2

4

6

3

3

3

3

3

3

3

3

2.5

2.5

2.5

2.5

2.5

2.5

2.5

2.5

2

2

2

2

2

2

2

2

1.5

1.5

1.5

1.5

1.5

1.5

1.5

1.5

1

1

1

1

1

1

1

1

0.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

0

0

0

0

0

0

0

0

−0.5

−0.5

−0.5

−0.5

−0.5

−0.5

−0.5

−0.5

0.8

0.8

0.8

0.8

0.8

0.8

6 1

1

1

1

1

5

4

3

2

Iteration − 1 1

0

2

4

6

0

2

4

6

−1

−1

−1

−1

−1

−1

−1

−1

−1.5

−1.5

−1.5

−1.5

−1.5

−1.5

−1.5

−1.5

−2

−2

−2

−2

−2

−2

−2

−2

0

5

0

5

0

5

0

5

0

5

0

5

0

5

0

5

ˆ Fig. 2 f1 (x) with mw = 0.01 and successive edge detection. Left: S(y). Right: edges found in each iteration marked by red cross symbols with the weak edges circled in green. Note that the ˆ after weak edges are almost invisible in the right figure of f1 (x). Each figure in the left shows S(y) each iteration from left to right. Each figure in the right shows the actual function and detected edges, from left to right

Figure 3 shows how the iteration method finds all edges. Figure 3 shows similar result as in Fig. 2. As shown in the figure, the method is highly accurate and finds all edges including the highly weak edges. As an application to the solution of PDEs, namely the shock-density wave interaction equation, we consider finding shocks in the density profile at t = 2 with the total number of grid points N = 300 computed with the WENO-Z method used in [9]. The left two figures of Fig. 4 show the edges (shocks) found by the Fourier concentration method while the right figure shows the edges (shocks) found by the iterative method. As shown in the figure, the iterative method find all the physical shocks accurately while the Fourier concentration method misses some of shocks. For two-dimensional examples, we consider a Shepp-Logan image with a faint box added to comprise additional weak edges, and a brain image. To detect edges in two dimensions, edges are detected slicewise in the x and y directions. The x and y coordinates have a range of [−π, π]. For a 2Nx +1×2Ny +1 image, slices of f (x, y) are taken at evenly spaced x and y with x = 2π/(2Nx + 1) and y = 2π/(2Ny + 1), with −π included and π excluded. Within each slice, Fourier coefficients are computed by partial Fourier expansion and the iterative method is applied to find strong and weak edges. Calculation parameters for the two-dimensional calculations were similar to those for the one-dimensional calculations. An edge with a concentration magnitude at or above a fraction τ = 0.1 of the maximum magnitude concentration was considered strong. To detect strong edges, trigonometric concentration factors with α = π were used. Figure 5 shows the edges found by the proposed method for the Shepp-Logan image. As in the figure, the weak edges (square box with magnitude of 0.01) are successfully found by the method. Remarks First, the proposed method is affected by noise as the original Fourier concentration. Consider Eq. (6). Let fˆk (mw = 0) be the Fourier coefficients with mw = mw  sin(7kπ/4) − sin(3kπ/2) 0. Then fˆk (mw ) = mw /8 for k = 0 and fˆk (mw ) = 2πk for k = 0. The weak edge translates fˆk . Thus we expect that unless SNR is high enough, |fˆk (mw = 0) − fˆk (mw )| becomes easily smaller than the noise as mw decays. Figure 6 shows the concentration with mw = 0 (left), the concentration with SNR = 20 (middle) and with SNR = 10. As in the figure, the weak edges are

1.5

1

0.5

0

−0.5

−1

−1.5

−2

1.5

1

0.5

0

−0.5

−1

−1.5

−2

5

0

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

3

5

0

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

3

5

0

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

3

5

0

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

3

5

0

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

3

5

0

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

3

5

Fig. 3 f1 (x) with mw = 0.001 and successive edge detection. Detected edges are marked by red cross symbols with the weak edges circled in green. Note that the edges in f1 (x) are almost invisible in the figure

0

2

2

5

2.5

2.5

0

3

3

Iterative Spectral Mollification and Conjugation for Successive Edge Detection 581

582

R. E. Tuzun and J.-H. Jung τ = 0.06

p=8 5

τ = 0.08

p=4

4

4

3

3

3

2

2

2

1 0 −1 0

1

τ = 0.06

1

2

ρ

4

ρ

ρ

Iterative Method 5

5

1

p=8

0

0

3 x

4

5

6

−1 0

−1 1

2

3 x

4

5

0

6

1

2

3 x

4

5

6

Fig. 4 The density ρ in y-axis versus the x-coordinate in x-axis at t = 1 for the shock-density wave interaction and shocks (with cross symbols) found with different values of τ and p. Left two figures: the Fourier concentration method. Right: the iterative method

20 40 60 80 100 120 140 160 180 200 20

40

60

80

100

120

140

160

180

200

Fig. 5 Original image (left) and edges (right) detected for concentration followed by iterative method on the Shepp-Logan image with a weak square edge of magnitude 0.01 added

0.4 0.2

Concentration

0.6

0 0

1

1 Concentration

Concentration

1 0.8

0.8 0.6 0.4

1

2

3 x

4

5

6

0.6 0.4 0.2

0.2 0

0.8

1

2

3 x

4

5

6

0 0

1

2

3 x

4

5

6

Fig. 6 Concentration in y-axis versus x. Left: without noise. Middle: SNR=20. Right: SNR=10

easily indistinguishable as SNR decreases. As the main objective of this research is finding the weak edges, a proper noise reduction suitable for the proposed method should be investigated in our future research. Second, the proposed method is to find

Iterative Spectral Mollification and Conjugation for Successive Edge Detection

10

0.14

8

0.12 0.1

6 df/dx

df/dx

Finite Difference of f(x)

×10-4

Finite difference of f(x) in physical domain

0.16

583

0.08 0.06

4

0.04

2

0.02

0

0 0

1

2

3

4

5

6

4.6

4.8

5

Fig. 7 Left: edges with finite difference in the physical domain. Right:

5.2

5.4

5.6

x

x

3π 2

≤x≤

7π 4

the edges by fˆk . Figure 7 shows the edge detection in the physical domain, with the forward difference, generated from the Fourier data (the left figure). The right figure 7π shows the plot in x ∈ 3π 2 ≤ x ≤ 4 where the weak edges exist. As in the figures, the weak edges are still hard to distinguish in the physical domain. Once the strong edges are removed in the Fourier domain and switching back and forth from the Fourier to physical domains, the weak edges are eventually found with the proposed method. Summary The following is the summary of the proposed iterative concentration method. The procedure stops eventually with a non-zero value of η > 0 in Eq. (10). • Step 1: Find edge locations xo using the Fourier concentration method. • Step 2: Apply the local filter near xo and find the new set of Fourier coefficients. • Step 3: Find a new edge location yo where the normalized concentration Sˆ by {fˆk } from Step 2 has the maximum. • Step 4: Repeat Steps 2 and 3 until all the edges are found (the iteration stops once all edges are found.)

4 Conclusion We showed that the iterative approach of the Fourier concentration method can detect all edges, which is not the case if the weak edges are too small. We showed that the proposed method is able to detect weak edges 3000 times weaker than the strongest edge, as long as the weak edges are well-separated from the stronger edges without noise and that the proposed method find all weak edges in a PDE application, namely the WENO calculation for the shock-density wave interaction. The iterative method also shows that it stops almost automatically after all the edges are found. Thus the proposed method is accurate and efficient.

584

R. E. Tuzun and J.-H. Jung

References 1. Archibald, R., Gelb, A.: Reducing the effects of noise in image reconstruction. J. Sci. Comput. 17, 167–179 (2002) 2. Cochran, D., Gelb, A., Wang, Y.: Edge detection from truncated Fourier data using spectral mollifiers. Adv. Comput. Math. 38, 737–762 (2013) 3. Don, W.-S., Wang, B.-S., Gao, Z.: Fast iterative adaptive multi-quadric radial basis function method for edges detection of piecewise functions–I: uniform mesh. J. Sci. Comput. 75, 1016– 1039 (2018) 4. Engelberg, S., Tadmor, E.: Recovery of edges from spectral data with noise—a new perspective. SIAM J. Numer. Anal. 46, 2620–2635 (2008) 5. Gelb, A., Tadmor, E.: Detection of edges in spectral data. Appl. Comput. Harmon. Anal. 7, 101–135 (1999) 6. Gelb, A., Tadmor, E.: Detection of edges in spectral data II: nonlinear enhancement. SIAM J. Numer. Anal. 38, 1389–1408 (2000) 7. Gelb, A., Tadmor, E.: Adaptive edge detectors for piecewise smooth data based on the minmod limiter. J. Sci. Comput. 28, 279–306 (2006) 8. Gottlieb, D., Tadmor, E.: Recovering pointwise values of discontinuous data within spectral accuracy. In: Murman, E.M., Abarbanel, S.S. (eds.). Progress and Supercomputing in Computational Fluid Dynamics. Proceedings of a 1984 U.S.-Israel Workshop, Progress in Scientific Computing, vol. 6, pp. 357–375. Birkhauser, Boston (1985) 9. Jung, J.-H., Durante, V.R.: An iterative adaptive multi quadric radial basis function method for the detection of local jump discontinuities. Appl. Numer. Math. 59, 1449–1466 (2009) 10. Jung, J.-H., Gottlieb, S., Kim, S.: Iterative adaptive RBF methods for detection of edges in two dimensional functions. Appl. Numer. Math. 61, 77–91 (2011) 11. Pang, J., Miller, E., Driban, J., Tassinari, A., McAlindon, T.: A curve evolution method for identifying weak edges with applications to the segmentation of magnetic resonance images of the knee. In: 2011 IEEE International Symposium on Biomedical Imaging: From Nano to Macro, pp. 1410–1415 (2011) 12. Petersen, A., Gelb, A., Eubank, R.: Hypothesis testing for Fourier based edge detection methods. J. Sci. Comput. 51, 608–630 (2012) 13. Stefan, W., Viswanathan, A., Gelb, A., Renaut, R.: Sparsity enforcing edge detection method for blurred and noisy Fourier data. J. Sci. Comput. 50, 536–556 (2012) 14. Tadmor, E., Tanner, J.: Adaptive mollifiers for high resolution recovery of piecewise smooth data from its spectral information. Found. Comput. Math. 2(2), 155–189 (2002) 15. Tadmor, E., Zou, J.: Three novel edge detection methods for incomplete and noisy spectral data. J. Fourier Anal. Appl. 14, 744–763 (2008) 16. Viswanathan, A., Gelb, A., Cochran, D.: Iterative design of concentration factors for jump detection. J. Sci. Comput. 51, 631–649 (2012)

Iterative Spectral Mollification and Conjugation for Successive Edge Detection

585

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Small Trees for High Order Whitney Elements Ana Alonso Rodríguez and Francesca Rapetti

1 Introduction We aim at determining in a constructive way, for the high order case, the finite element solutions of grad φ = E, curl A = B, div D = ρ, namely, of the equations linking the electric field E, the magnetic induction B, and the electric charge density ρ, to their potentials φ, A and D, respectively. Stating the necessary and sufficient conditions for assuring that a function defined in a bounded set Ω ⊂ R3 is the gradient of a scalar potential, the curl of a vector potential or the divergence of a vector field is one of the most classical problem of vector analysis (see for example [3, 6, 8]). We aim at providing an explicit and efficient procedure to construct a finite element solution. For example, div-free fields, W, are implicitly characterized in terms of a vector w of degrees of freedom of W by the algebraic constraint D w = 0, with D the matrix of the div operator between finite elements spaces. The same fields, in the case of a domain with connected boundary, are explicitly defined by w = R a, with no constraint on a, where R is the matrix of the curl operator between finite elements spaces and a collects the degrees of freedom of the vector potentials A. Similarly, one can wish to compute a vector potential a such that R a = b, for a given field b verifying D b = 0. As explained in [5], these bases can be constructed by the help of “trees” and “co-trees”, which are at the core of this contribution. The case r = 0 is largely treated in the literature for different types of topological domains (see for example [2]). In these pages, we develop the

A. Alonso Rodríguez () Dipartimento di Matematica, Università degli Studi di Trento, Trento, Italy e-mail: [email protected] F. Rapetti Département de Mathématiques, Université Côte d’Azur, Nice, France e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_47

587

588

A. Alonso Rodríguez and F. Rapetti

tree and co-tree approaches for r > 0 when fields in the high order Whitney spaces are represented on the basis of their weights on small simplices [7, 9, 10]. With this choice of degrees of freedom, the tree and co-tree concepts extend from r = 0 to r > 0 straightforwardly.

2 Basic Concepts Let Ω ⊂ R3 be a bounded polyhedral domain with Lipschitz boundary ∂Ω and ¯ We denote by |A| the cardinality of the set A. For M a simplicial mesh of Ω. 0 ≤ k ≤ 3, let Δk (T ) (resp. Δk (M)) be the set of k-simplices of a mesh tetrahedron T (resp. of the mesh M). Note that Δk (M) = ∪T ∈M Δk (T ). If Δ0 (M) = {vi }i , with i = 1, . . . , Nv , being Nv = |Δ0 (M)|, then each k-simplex S ∈ Δk (M) has associated an increasing map mS : {0, . . . , k} → {1, . . . , Nv }. This map induces an (inner) orientation on S (i.e., a way to run along S if k = 1, through S if k = 2, in S if k = 3). If < we assign to each S ∈ Δk (M) a real number cS we can define the k-chain c = S∈Δ0 (M) cS S, i.e. a formal < weighted sum of k-simplices S in M. One can add k-chains, namely (c + c) ˜ = S (cS + c˜S ) S, and multiply a k-chain by a scalar < p, namely p c = S (p cS ) S. The set of all k-chains in M, here denoted Ck (M), is a vector space, in one-to-one correspondence with the set of real vectors c = (cS )S∈Δk (M) . Each k-simplex S ∈ Δk (M), can be associated with the elementary k-chain c with entries cS = 1 and cS˜ = 0 for S˜ = S. In the following we will use the same symbol S to denote the oriented k-simplex and the associated elementary k-chain. The boundary operator ∂ takes a k-simplex S and returns the sum of all its (k − 1)-faces f with coefficient 1 or −1 depending of whether the orientation of the (k − 1)-face f matches or not with the orientation induced by that of the simplex S on f . Since the boundary operator is a linear mapping from Ck (M) to Ck−1 (M), it can be represented by a matrix ∂ of dimension |Δk−1 (M)|×|Δk (M)|, which is rather sparse, gathering the coefficients 0, −1, or +1. Note that in three dimensions, there are three nontrivial boundary operators acting, respectively, on edges, triangles and tetrahedra: ∂1 represented by the matrix G8 , ∂2 represented by R8 , and ∂3 represented by D8 . To fully specify ∂, we need to specify the boundary of each simplex S. By definition, we have ∂1 e =

 n∈Δ0 (M)

Ge,n n,

∂2 f =

 e∈Δ1 (M)

Rf,e e,

∂3 T =



DT ,f f,

f ∈Δ2 (M)

for any e ∈ Δ1 (M), any f ∈ Δ2 (M) and any T ∈ Δ3 (M). For e = [v0 , v1 ], f = [v0 , v1 , v2 ] and T = [v0 , v1 , v2 , v3 ], we have, respectively, ∂2 [v0 , v1 , v2 ] = [v0 , v1 ] − [v0 , v2 ] + [v1 , v2 ], ∂1 [v0 , v1 ] = v0 − v1 , ∂3 [v0 , v1 , v2 , v3 ] = [v0 , v1 , v2 ] − [v0 , v1 , v3 ] + [v0 , v2 , v3 ] − [v1 , v2 , v3 ].

Small Trees for High Order Whitney Elements

589

The subscript is removed when there is no ambiguity, since the operator needed for a particular operation is indicated from the type of the operand (e.g., ∂3 when ∂ applies to tetrahedra). The notion of < ;0, the exterior derivative of the (k − 1)-form w is the k-form dw such that s dw = ∂s w for all s ∈ Ck (M). With this simple equation relating the evaluation of dw on a simplex s to the evaluation of w on the boundary of this simplex, the exterior derivative is readily defined. We can naturally extend of a differential form w on an arbitrary ; the notion of ; 1 can be visualized starting from the principal lattice Lr+1 (T ) in the simplex T = {nσ 0 (0) nσ 0 (1) nσ 0 (2) nσ 0 (3) } defined as T

 Lr+1 (T ) = x ∈ T : λσ 0 (i) (x) ∈ {0, T

T

T

T

 2 r 1 , ,..., , 1}, 0 ≤ i ≤ 3 . r +1 r +1 r +1

and connecting its points by edges parallel to those of T . (See, e.g., Fig. 1.) We denote by Λk (Ω) the space of all smooth differential k-forms on Ω. The completion of Λk (Ω) in the corresponding norm defines the Hilbert space − L2 Λk (Ω). Let Pr+1 Λk (T ) be the space of so-called trimmed polynomial k-forms of degree r + 1 on T , with r ≥ 0, (as in [7]), and we define − − Pr+1 Λk (M) = {ω ∈ H Λk (Ω) : ω|T ∈ Pr+1 Λk (T ), T ∈ M}

where H Λk (Ω) = {ω ∈ Λk (Ω) : dω ∈ Λk (Ω)} is a Hilbert space (see [4]). − Definition 1 The weights of a polynomial k-form u ∈ Pr+1 Λk (T ), with 0 ≤ k ≤ 3 and r ≥ 0, are the scalar quantities  u, (1) {α,S}

on the small simplices {α, S} with α ∈ I(4, r) and S ∈ Δk (T ).

v3 v3

v3

v2 v0

v0

v0 v1

v2

v2

v1

v1

Fig. 1 From the principal lattice of degree r +1 = 3 in a tetrahedron T , we define a decomposition of T into 10 small tetrahedra, 4 octahedra O and 1 reversed tetrahedron. Each face on ∂T is decomposed into 6 small faces and 3 reversed triangles, in solid red line (Left)

Small Trees for High Order Whitney Elements

591

We now list some remarkable properties of the small simplices which are useful in the tree construction. − Property 1 The weights (1) of a Whitney k-form u ∈ Pr+1 Λk (T ) on all the small simplex {α, S} of T are unisolvent, as stated in [7, Proposition 3.14]. The small − simplices can thus support the degrees of freedom for fields u ∈ Pr+1 Λk (T ), with 0 ≤ k ≤ 3 and r ≥ 0. Since the result on unisolvence holds true also by replacing − T with F ∈ Δn−1 (T ) then TrF u ∈ Pr+1 Λk (F ) is uniquely determined by the weights on small simplices in F . It thus follows that a locally defined u, with u|T ∈ − Pr+1 Λk (T ) and single-valued weights, is in H Λk (Ω). We thus can use the weights on the small simplices {α, S} as degrees of freedom for the fields in the finite element − space Pr+1 Λk (M) being aware that their number is greater than the dimension of the space.

Property 2 The weights given in Definition 1 have a meaning as cochains and this relates directly the matrix describing the exterior derivative;with the;matrix of the boundary operator. The key point is the Stokes’ theorem C du = ∂C u , where − Λk (M) then u is a (k − 1)-form and C a k-chain. More precisely, if u ∈ Pr+1 − z = du ∈ Pr+1 Λk+1 (M) and 

 {α,S}

z=

 {α,S}

du =

u= ∂{α,S}

 {β,F }

 B{α,S},{β,F }

{β,F }

u

being B the boundary matrix with as many rows as small simplices of dimension k and as many columns as small simplices of dimension k − 1. The small simplices {α, S} inherit the orientation of the simplex S so the coefficient B{α,S},{β,F } is equal to the coefficient BS,F of the boundary of the simplex S if β = α. This is straightforward if dim(F ) > 0 and when dim(F ) = 0, providing that small nodes in T are given in the notation {α, n} according to their position in the small simplices when fragmented (see Fig. 1 in [1]). Property 3 The generated (r+2 2 ) small faces on each face F of T , pave F together r+1 with the ( 2 ) reversed triangles, denoted by ∇, contained in F . Similarly, the r+2 generated (r+3 3 ) small tetrahedra contained in T pave T together with the ( 3 ) octahedra, denoted by O, and the (r+1 3 ) reversed tetrahedra, denoted by ⊥, contained in T , as shown in Fig. 1. Reversed octahedra and reversed tetrahedra are examples of “holes” in T (see [9, 10]). Property 4 Since homology is preserved by homotopy, in [10, Section 3.4], it is discussed the fact that the relative homology (i.e., the homology [modulo the holes’ boundaries]), of the complex of small simplices is the same of the homology of M. This property is fundamental to build the tree for high order potentials when working with small simplices. The homology [modulo the holes’ boundaries] can be translated in matrix notation, by showing that the boundary matrices associated with

592

A. Alonso Rodríguez and F. Rapetti

the small simplices, “modified” and “completed” (in a sense that we explain in the next section) by the relations [10, Proposition 3.5] are incidence matrices of a graph. To apply the theory presented in [10, Section 3.4] in a tetrahedron T ∈ Δ3 (M), we need to introduce, for r > 0, two sets K1 and K2 of chains generated by the small simplices that belong to the boundary of some hole in T as follows: • K1 are the chains generated by the boundary of the (r+1 2 ) reversed triangle ∇ ⊂ F and that for each F ∈ Δ2 (T ), and the boundary of the three faces out of four on the boundary ∂⊥ of each of the (r2 ) reversed tetrahedra ⊥ in T ; • K2 are the chains generated by 4 out of 8 faces of the (r+2 3 ) octahedra O in T . The involved faces are the small faces belonging to the boundary ∂O privated of ∂O ∩ (Δ2 (T ) ∪ ∂⊥). The two sets K1 and K2 satisfy the property ∂K2 ⊂ K1 , decisive to conclude that the relative homology [modulo the holes’ boundaries] of the complex of the small simplices is the same as the homology of the original mesh M [10].

4 Trees and Graphs As stated in [12], a directed graph G consists of two sets N and A of nodes and arcs, respectively, subjected to certain incidence relations, collected in the all-vertex incidence matrix MG ∈ Z|N |×|A| as follows: ⎧ ⎪ ⎨ −1 , if a starts from n, MG = +1 , if a ends in n, n,a ⎪ ⎩ 0 , if a does not contain n. An incidence matrix M of the graph G is any sub-matrix of MG with |N | − 1 rows and |A| columns. The node that corresponds to the row of MG that is not in M will be indicated as the reference node of G. A graph G is connected if there is a path between any two of its nodes. A tree T of a graph G is a connected acyclic subgraph of G. A spanning tree Ts is a tree of G visiting all its nodes. Any connected graph G admits a spanning tree Ts . We have now to particularize these notions for small simplices. In each tetrahedron T of the oriented mesh M, we consider the small mesh associated with Lr+1 (T ) composed only of small tetrahedra, for a given r uniform all over the mesh M. The union of the small meshes for all T ∈ Δ3 (M) is denoted Mall . A (Primal) Small Tree for the Gradient Problem For r = 0, the graph G 1 has N = Δ0 (M) and A = Δ1 (M). The boundary matrix G8 is the all-vertex incidence matrix of the graph G 1 . Extracting a spanning 1-tree Ts1 from G 1 is equivalent to finding in G8 , minus one row, a submatrix of maximal rank (see [11] for a suitable and easy way of constructing T ). For r > 0, we have

Small Trees for High Order Whitney Elements

593

v0

v0 v3 v1

v2

v3

v0 v3 v1

v2

v1 v2

Fig. 2 (Left) The graph G 1 and a spanning tree in thick line, for r = 1. (Right) A spanning tree for r = 2 in a fragmented layout

to consider the new graph G 1 with N = Δ0 (Mall ) and A = Δ1 (Mall ). Let G8 all be 8 1 the all-vertex incidence matrix of this new graph G . Note that Gall results from the boundary operator ∂1 on the elementary 1-chains from Mall . Extracting a spanning 1-tree Ts1 from G 1 is equivalent to finding in G8 all , minus one row, a submatrix of maximal rank. Example of spanning 1-tree Ts1 for r +1 = 2 in the right part of Fig. 2 and for r + 1 = 3 in Fig. 5 (fragmented visualization). Note that we can repeat this construction in the two-dimensional case. A (Dual) Small Tree for the Divergence Problem For r = 0, the graph G 2 is built on M∗ , the so-called dual mesh of M, as follows. Let us note that an internal face F ∈ Δ2 (M) connects two adjacent tetrahedra T1 , T2 ∈ Δ3 (M) whereas a boundary face Fb ∈ Δ2 (M) connects a tetrahedron Tb ∈ Δ3 (M) and the boundary ∂Ω. We can construct the following connected (dual) graph G 2 : the set of nodes, N , contains the barycenter of any tetrahedron T ∈ Δ3 (M) together with one additional exterior node representing ∂Ω; the set of arcs, A, contains any face F ∈ Δ2 (M). For r = 0, the matrix D associated with the boundary operator ∂3 , acting on C3 (M), is an incidence matrix of the (dual) graph G 2 , with reference node the one corresponding to ∂Ω. Extracting a spanning tree Ts2 from G 2 is equivalent to finding in D a submatrix of maximal rank. For r > 0, let R2 be the set of small faces chosen as follows: one small face for each octahedron O contained in K2 (see the right side of Fig. 3 for the dashed small face in R2 when r +1 = 2). To construct the graph G 2 for r > 0 we need to consider M∗all , the dual mesh associated to Mall , where nodes are the small tetrahedra and the arcs the small faces, apart from the ones in R2 . To understand this, we can reason

594

A. Alonso Rodríguez and F. Rapetti

v0

v0 3 arcs

3 arcs

v3

v1 3 arcs

v3

v1 3 arcs

v2

v2

3 arcs

3 arcs

Fig. 3 The (dual) graph G∗2 associated with the small mesh Mall defined in a tetrahedron T for r = 1: the black dots are the nodes, and curved lines the arcs (Left). The (dual) graph G 2 obtained from G∗2 by merging the nodes corresponding to barycenter of t0 = {(1, 0, 0, 0), T } and of O, thus eliminating the arc associated with the shaded small face fuO (Right)

as follows. For r > 0, we have one arc connecting two small tetrahedra, say t# , t◦ , when • either t# , t◦ share the same small face f , i.e. ∂t# ∩ ∂t◦ = f ; • or t# , t◦ have a small face on the boundary of the same octahedron O, i.e. f# = ∂t# ∩ ∂O and f◦ = ∂t◦ ∩ ∂O for the same octahedron O. See an example of graph G 2 for Mall (here M = {T }) in the left part of Fig. 3 for r + 1 = 2, where the node associated with the octahedron O is not a node in the graph, but stands to indicate that the four small tetrahedra are connected one to the other by one arc because they all have one small face on ∂O. Naming tk the small tetra with a vertex in vk , k = 0, 3, and numbering first the 3 × 4 faces on tk ∩ ∂T , called fik for i = 1, 2, 3, second those on ∂O (where fuO , fO , fdO , frO are the small faces up, left, down, right of ∂O), we have ⎛ ∂T

Dt mp

−1 −1 −1 −1 −1 −1 −1 −1 −1 −1 −1 −1

⎜ t0 ⎜ ⎜ 1 ⎜ t1 ⎜ ⎜ = t2 ⎜ ⎜ ⎜ t3 ⎜ ⎝ O

1

⎞ ⎟ ⎟ ⎟ ⎟ ⎟ −1 ⎟ ⎟ −1 ⎟ ⎟ −1 ⎟ ⎠ 1 1 1 1

−1

1 1

1

1 1

1

1 1

1

1

f10 f20 f30 f11 f21 f31 f12 f22 f32 f13 f23 f33 fuO fO fdO frO Since the octahedron O is not part of the small mesh Mall , we have to imagine that its node collapses with the node of one of its neighbouring small tetrahedron, say t0 with a vertex in v0 , and thus that the corresponding arc (i.e. the small face fuO = ∂t0 ∩ ∂O, the dashed one in the right part of Fig. 3) is eliminated. From a

Small Trees for High Order Whitney Elements

595

v0

v0

v3

v3

v1

v1

v2

v2

Fig. 4 Example of spanning tree in the (dual) graph G 2 , namely a selection of acyclic paths made of arcs, visiting all the nodes of G 2 (r = 1, Left and r = 2, Right)

matrix point of view, D is obtained by adding the line “O” in Dt mp to the line “t0 ”, and eliminating fuO , namely t0 t1 D = t2 t3

⎛ ⎜ ⎜ ⎜ ⎝

1

1

1 1

1

1 −1

1 1

1

1

f10 f20 f30 f11 f21 f31 f12 f22 f32

1

−1 1 1 1 f13 f23 f33 fO fdO frO

1

⎞

⎟ ⎟ ⎟ ⎠ −1

(in bold font, the submatrix of maximal rank in D for the spanning tree Ts2 illustrated in Fig. 4, left part for r + 1 = 2). To repeat this construction in the two-dimensional case, when T is a triangle, we have to consider the mesh Mall of small triangles in T and the role of the core octahedra O is played by the reversed triangles ∇ ∈ T . The set R2 is replaced by R1 , composed of one small edge for each reversed triangle ∇ ∈ K1 . In two dimensions we do not have reversed tetrahedra, therefore no reversed triangles ∇⊥ . The construction of the spanning tree in Mall can be done by assembling that of the geometrical mesh M, namely a spanning tree for the Whitney forms of lower degree (blue lines in Fig. 5 (Right)), together with local contributions, one from each element (green lines in Fig. 5 (Right)). Each local contribution results from one fixed on a reference element which is mapped on the current element (respecting the orientation). In Fig. 5 (Left), in green/red thick line we have marked the small edges of a spanning tree in the graph G 1 , for r = 3, in the reference triangle. The red ones belong to the spanning tree in the reference triangle, but they are in general omitted in the spanning tree of Mall , (indeed, they appear only if they are covered by the

596

A. Alonso Rodríguez and F. Rapetti 5

3

0 6

4

1

2

7

2

1

Fig. 5 (Left) In thick colored line, the small edges of the graph G 1 , for r = 3, that compose a spanning tree in a reference triangle. (Right) In thick blue line the contribution of the branches of a spanning tree in a (2D) toy mesh M reported on Mall . In green, the contribution of the small branches mapped from the green ones in the reference triangle. It is not necessary to report the red ones since they are either covered by the blue ones or omitted. The co-tree is in black

blue tree). The small co-tree is in black. A similar construction can be repeated in 3D (both for k = 1 and k = 2) and it reflects the decomposition given, for instance, in [13] (Sect. 5).

References 1. Alonso Rodríguez, A., Rapetti, F.: The discrete relations between fields and potentials with high order Whitney forms. In: Radu, F.A., et al. (eds.) European Conference on Numerical Mathematics and Advanced Applications. Enumath 2017. Lecture Notes in Computational Science and Engineering LNCSE, vol. 126. Springer, Berlin (2018) 2. Alonso Rodríguez, A., Valli, A.: Finite element potentials. Appl. Numer. Math. 95, 2–14 (2015) 3. Amrouche, C., Bernardi, C., Dauge, M., Girault, V.: Vector potentials in three-dimensional nonsmooth domains. Math. Methods Appl. Sci. 21, 823–864 (1998) 4. Arnold, D., Falk, R., Winther, R.: Finite element exterior calculus, homological techniques, and applications. Acta Numer. 15, 1–155 (2006) 5. Bossavit, A.: Computational Electromagnetism: Variational Formulations, Complementarity, Edge Elements. Academic, New York (1998) 6. Cantarella, J., DeTurck, D., Gluck, H.: Vector calculus and the topology of domains in 3-space. Am. Math. Mon. 109, 409–442 (2002) 7. Christiansen, S.H., Rapetti, F.: On high order finite element spaces of differential forms. Math. Comput. 85/298, 517–548 (2016) 8. Girault, V., Raviart, P.A.: Finite Element Methods for Navier-Stokes Equations: Theory and Applications. Springer, New York (1986) 9. Rapetti, F., Bossavit, A.: Geometrical localization of the degrees of freedom for Whitney elements of higher order. IEE Proc. Science, Meas. Technol. 1/1, 63–66 (2007); Special Issue on “Computational Electromagnetism” 10. Rapetti, F., Bossavit, A.: Whitney forms of higher degree. SIAM Numer. Anal. 47/3, 2369– 2386 (2009) 11. Rapetti, F., Dubois, F., Bossavit, A.: Discrete vector potentials for non-simply connected threedimensional domains. SIAM Numer. Anal. 41/4, 1505–1527 (2003)

Small Trees for High Order Whitney Elements

597

12. Stillwell, J.: Classical Topology and Combinatorial Group Theory. Springer, New-York (1993) 13. Zaglmayr, S.: High Order Finite Element Methods for Electromagnetic Field Computation. Ph.D. Thesis, Johannes Kepler Universität Linz, 2006

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Non-conforming Elements in Nek5000: Pressure Preconditioning and Parallel Performance A. Peplinski, N. Offermans, P. F. Fischer, and P. Schlatter

1 Introduction One of the most important concerns when solving numerically partial differential equations is finding the optimal grid on which the solution will be computed. Unfortunately in most cases it is not an easy task that could be determined in advance without deep understanding of the studied problem. That is why selfadapting algorithms like e.g. adaptive mesh refinement (AMR) have received much attention in past decades and became an important part of many packages for numerical modelling of fluid dynamics e.g. [9, 18]. The goal of AMR is to control the computational error during the simulation by placing higher resolution grids where it is needed. This makes the numerical modelling more robust, and gives the possibility to increase the accuracy of numerical simulations at minimal computational cost. The drawback is, however, increased solver complexity, and it that can have negative effects on the parallel code performance, in particular related to load balancing. There are number of different AMR schemes, and in the context of the spectral element method (SEM) [16], in which the discretisation is based on a decomposition of the computational domain into a number of non-overlapping, high-order subdomains called elements, we can distinguish three different categories: The mesh adaptation in this case can mean adjusting the (local) size of an element (r-

A. Peplinski () · N. Offermans · P. Schlatter Linné FLOW Centre and Swedish e-Science Research Centre (SeRC), KTH Mechanics, Stockholm, Sweden e-mail: [email protected]; [email protected]; [email protected] P. F. Fischer CS and MechSE Depts., University of Illinois at Urbana–Champaign, Champaign, IL, USA e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_48

599

600

A. Peplinski et al.

refinement), changing the polynomial order in a particular element (p-refinement), or splitting the element into smaller ones (h-refinement). In this work we concentrate on an h-refinement framework and its implementation in Nek5000 [8], which is a highly parallel and efficient SEM solver for the incompressible Navier–Stokes equations. In its established version, Nek5000 only supports conformal elements at constant polynomial order throughout the domain. The present work was started within EU project CRESTA, where the nonconforming solver for advection-diffusion problem was developed and the basic AMR tasks were implemented using existing external libraries. As h-refinement affects the element connectivity resulting in non-conforming meshes, a special grid manager is required to perform local refinement/coarsening and to build globally consistent meshes. For this task the p4est library [1] has been chosen, as it is designed to manipulate domains composed of multiple, non-overlapping logical cubic sub-domains, which can be represented by a recursive tree structure. This library provides element connectivity information for the dual graph, which is later manipulated by ParMETIS [10] producing a new element-to-processor mapping. The final step of grid refinement/coarsening and redistribution is performed within the non-conforming version of Nek5000, which utilises the so-called conformingspace/nonconforming-mesh approach based on the previous work of Fischer et al. [7, 11]. As the solver complexity grows special care has been taken to develop efficient tools that can be used within AMR framework. A more detailed description of them and the related scaling tests can be found in [17]. The goal of ExaFLOW is to extend results of CRESTA to the full incompressible Navier–Stokes equations focusing on proper adaptation of the pressure preconditioners for nonconforming SEM. Defining a robust parallel preconditioning strategy has received much attention in past decades, as the linear sub-problem associated with the divergence-free constraint (pressure-Poisson equation) can become very illconditioned. In the context of SEM two possible approaches based on the additive overlapping Schwarz method [4, 6] and the hybrid Schwarz-multigrid method [5, 12] were proposed and implemented in Nek5000, leading to a significant reduction of pressure iterations. In the present paper, we discuss the modifications necessary to adapt Nek5000 for the h-type AMR framework. The article is organised as follows. A short description of SEM and pressure preconditioners is given in Sects. 2 and 3. The following Sects. 4 and 5 describe the algorithmic modifications and parallel performance of the code. Finally, Sect. 6 provides conclusion and future work.

2 SEM Discretisation of the Navier–Stokes Equations We review briefly the discretisation of the incompressible Navier–Stokes equations to introduce notation and point out algorithm parts that require modification. The more in-depth derivation can be found in e.g. [4]. The temporal discretisation is based on a semi-implicit scheme in which the nonlinear term is treated explicitly

Non-conforming Elements in Nek5000

601

and the remaining unsteady Stokes problem is solved implicitly. To avoid spurious pressure modes our spatial discretisation is based on the PN − PN−2 SEM, where velocity and pressure spaces are spanned by Lagrangian interpolants on the Gauss–Lobatto–Legendre (GLL) and Gauss–Legendre (GL) quadrature points, respectively. Note that the basis for velocity is continuous across element interfaces, whereas the basis for pressure is not. Assuming fn incorporates all nonlinear and source terms treated explicitly at time t n , the matrix form of the Stokes problem after applying the Uzawa decoupling reads: 9

−1 T H − t β0 HB D 0 E

:#

un

p

$

# =

Bfn + DT pn−1 g

$ ,

(1)

where E=

t DB−1 DT β0

(2)

is the Stokes Schur complement governing the pressure, p = pn − pn−1 is the pressure update, and g is the inhomogeneity arising from Gaussian elimination. β0 1 In these equations H = − Re A + t B and D are the discrete Helmholtz and divergence operators, respectively. β0 , A and B denote here a coefficient from time derivative, a discrete Laplacian and a diagonal mass matrix associated with the velocity mesh. Applying the Uzawa decoupling we use the inverse mass matrix B−1 as approximation of the inverse Helmholtz operator H−1 , giving rise to a splitting error. Note that for this splitting method the diagonality of the mass matrix B is crucial to avoid costly matrix inversion. All operators H, A, B and E are symmetric positive definite (SPD) and can be solved with a preconditioned conjugate gradient (PCG) method. Moreover, E has properties similar to a Poisson operator, and is often referred to as a consistent Poisson operator. The systems involving H and E are solved iteratively with E being more challenging, and in the next section we will present the preconditioning strategy for the pressure equation, E p = −Du .

(3)

We close this section by shortly presenting the SEM operators. SEM introduces a globally unstructured and locally structured basis by tessellating the domain into 8K K non-overlapping subdomains (deformed quadrilaterals),  = k=1 k , and representing functions in each subdomain in terms of tensor-product polynomials ˆ = [−1, 1]d . In this approach every function or operator on a reference subdomain  is represented by its local counterparts, which in case of functions takes the form of a sum over the subdomains f (x) =

K   k=1

i

fik hi (r) .

602

A. Peplinski et al.

Here, fik and hi are the nodal values of the function in k and the base functions ˆ respectively, with i representing the natural ordering of nodes in . ˆ Comin , k bining the coefficients fi one can build global f and local f L representations of the function. Each global degree of freedom occurs only once in the global representation, but has multiple copies of faces, edges and vertices related to k in the local one. To enforce function continuity, the global-to-local mapping is defined as the matrix–vector product f L = Qf , where Q is a binary operator duplicating the basis coefficients in adjoining subdomains. The action QT f L sums multiple contributions to the global degree of freedom from their local values. The assembled global stiffness matrix A takes the form 

 ∇f, ∇g = f T Ag = f T QT AL Qg,

where a block diagonal matrix AL is the unassembled stiffness matrix with each ; dhi dhj diagonal block consisting of the local stiffness matrix Akij = dx dx dx. In practise, the global stiffness matrix is never formed explicitly, and the gather–scatter operator QQT is used instead. This operator contains all information about element connectivity.

3 Pressure Preconditioner An efficient solution of Eq. (3) requires finding an SPD preconditioning matrix M−1 which can be inexpensively applied and which reduces the condition number of M−1 E. Preconditioners based on domain decomposition are a natural choice for SEM as the data is structured within an element but is otherwise unstructured. An overlapping additive Schwarz preconditioner for Eq. (3) was developed in [4] based on linear finite element discretisation of Poisson operator. It combines solutions of the local Poisson problems in overlapping subdomains RTk Aˆ −1 k Rk with −1 T ˆ the coarse grid problem R0 A0 R0 , which is solved on few degrees of freedom, but covers the entire domain  RTk Aˆ −1 M−1 = RT0 Aˆ −1 k Rk . 0 R0 + k

For the local problems restriction and prolongation operators, Rk and RTk , are ˆ k is a local Boolean matrices that transfer data to and from the subdomain, and A stiffness matrix which can be inverted with e.g. a fast diagonalisation method. Note that action of Rk and RTk are similar to the gather–scatter operator QQT . The coarse grid problem corresponds to the Poisson problem solved on the element vertices only, with RT0 being the linear operator interpolating the coarse grid solution onto the tensor product array of GL points. Unlike in [4, 6], Aˆ 0 is

Non-conforming Elements in Nek5000

603

defined using local SEM-based Neumann operators performing the projection of local stiffness matrices Ak evaluated on the GLL quadrature points onto the set of coarse base functions bi representing the linear finite element base on the GLL ˆ as a tensor-product of the onegrid. The coarse base functions are defined in  dimensional linear functions. The local contribution to Aˆ 0 is given by biT Ak bj , and the full Aˆ 0 is finally assembled by local-to-global mapping summing contributions to the global degree of freedom from their local counterparts. Aˆ 0 is one of few matrices formed explicitly in Nek5000. On the other hand, the hybrid Schwarz-multigrid preconditioner is based on the multiplicative Schwarz method, which for the two-level scheme takes the form, ⎡ ⎣ M−1 = RT0 Aˆ −1 0 R0



⎤ ⎦ RTk Aˆ −1 k Rk ,

k

and leads to the following two-level multigrid scheme, (i) u1 =



RTk Aˆ −1 k Rk g,

k

(ii) r = g − Au1 , ˆ −1 R0 r, (iii) e = RT0 A 0 (iv) u = u1 + e, where g, r, e and u are right-hand side, residual, coarse-grid error and solution of equation Au = g, respectively. This method can be extended to a general multilevel solver performing a full V cycle [5, 12]. Notice that by replacing step ii) with r = g we obtain the additive Schwarz preconditioner.

4 Adaptation for Non-conforming Meshes The important advantage of SEM in the context of AMR is its spatial decomposition into elements that can easily be split into smaller ones, and use of the local representation of the operators which decouples intra- and inter-element operations. As h-type AMR using the conforming-space/nonconforming-mesh approach leaves the approximation spaces unchanged, most of the tensor-product operations evaluated element-by-element are preserved, limiting the changes in the algorithm. The inter-element operations are mostly performed by the gather–scatter operator QQT which has to be redefined to include spectral interpolation at the nonconforming faces. Following [7] we consider a non-conforming face shared by one low resolution element (parent) and two (in 3D four) high resolution elements

604

A. Peplinski et al.

(children). We introduce a local parent-to-child interpolation operator Jcp which is a spectral interpolation operator with entries  cp  cp J ij = hj (ζi ), cp

where ζj represents the mapping of GLL points from the child face to its parent. This operator is locally applied to give the desired nodal values on the child face, after Q copies data form the parent to the children. Building a block-diagonal matrix JL with local matrices Jcp one can redefine scatter JL Q and gather–scatter JL QQT JTL operators, respectively. For more discussion see Fig. 6 and Sect. 4 in [7]. The next crucial modification is diagonalisation of the global mass matrix QT BL Q (BL is a block-diagonal built of local mass matrices), whose inverse is required in Eqs. (1) and (2). It is non-diagonal due to the fact that the quadrature points in the elements along the non-conforming faces do not coincide. A diagonalisation procedure is given in [7] and consists of building the global vector b˜ b˜ := Bˆe = QT JTL BL eˆL , and finally setting the lumped mass matrix B˜ ij = δij b˜ i . eˆ and eˆ L denote here the global and local vectors containing all ones. The additive Schwarz preconditioner requires two significant modifications. The first one is related to the assembly of the coarse grid operator Aˆ 0 , which gets more complex for non-conforming meshes. This is due to the fact that the non-conforming mesh introduces hanging vertices located in the middle of faces or edges. These hanging vertices are not global degrees of freedom and cannot be included in Aˆ 0 . To remove them from consideration one has to modify the set of local coarse base functions bi , which are thus dependent on the shape of the refined region as well as the position and orientation of the child face with respect to the parent one. Unlike the conforming case, where all bi could be represented by a tensor product of two or three linear functions, the non-conforming mesh requires 5 basic components in two and 21 in three dimensions to assemble all the possible shapes of bi . The last missing components are the restriction and prolongation operators, Rk and RTk , for the local Poisson problem. Taking into account the similarity between these operators with QQT and following the previous development we use an operator similar to JL QQT JTL , replacing JL with the interpolation operator defined on the GL quadrature points. Although this choice seems to be optimal as it preserves properties of the preconditioner and JTL is well defined, our numerical experiments showed a significant increase of pressure iterations in some cases. It was found to be caused by the noise introduced by JTL in the Schwarz operator. To reduce this noise we replaced the transposed interpolation operator with the inverse one, getting a significant reduction of iterations. Unfortunately, such a preconditioner is no longer SPD and PCG cannot be used as an iterative solver cp in this case. The other problem is the definition of J−1 L , as J can be inverted for

Non-conforming Elements in Nek5000

605

square matrices only, thus excluding p-refinement strategies. To avoid this problem we define a child-to-parent interpolation operator Jpc with the entries  pc  J ij =

+

pc

p

hj (ζi ) if ζj ∈ ∂p ∩ ∂c , 0 otherwise p

where ∂p and ∂c are the parent and child common faces, ζj is a parent GLL pc p point at the face ∂p , and ζj represents the mapping of ζj to the child face ∂c . This operator is locally applied to give the desired nodal values on the child face, before QT sums data form the children and the parent. Building block-diagonal pc one can redefine the gather–scatter matrix J−1 L consisting of local matrices J T −1 JL QQ JL operator such that it is appropriate for the pressure preconditioner. In a similar way we modify the multiplicative Schwarz method, as it shares a number of features with the additive one. In this case we distinguish between Schwarz (acting at single level) and restriction (connecting different levels) operaT T tors and apply JL QQT J−1 L and JL QQ JL to each of them, respectively. Unlike the additive preconditioner, the hybrid one requires also the redefinition of the diagonal weight matrix that indicates the number of sub-domains sharing a given node, and is used to accommodate for overlapping regions. Its value is important as it reduces the largest eigenvalue of the MA operator and defines the smoothing properties of the additive Schwarz step (see [4] and the references therein). In the conforming case its definition is straightforward, however the non-conforming case is more involved as hanging nodes are not real degrees of freedom. In the current implementation the information about node multiplicity on the non-conforming faces is hidden to the parent element, so the parent element sees only one neighbour instead of two (four in 3D). Although this choice gives a preconditioner that significantly reduces the number of pressure iterations, its performance for the studied cases is slightly worse than the performance of the additive Schwarz preconditioner. This can be caused by a non-optimal value of the weight matrix, or by the fact that the hybrid preconditioner is superior over the additive one for high-aspect ratio elements (that are not present in our adaptive simulations).

5 Parallel Performance The parallel performance test is based on the one of the ExaFLOW flagship calculations, and consists of the turbulent flow around a NACA4412 wing section with 5◦ angle of attack, at a Reynolds number based on inflow velocity U∞ and chord length c of Rec = 200,000. It was previously studied in a series of wellresolved large-eddy simulations conducted with the conforming Nek5000 version, and discussed in detail in [19]. This flow configuration was chosen to illustrate the significant benefit of using AMR, in particular when it comes to the farfield region in the computational domain, but for this article we will only briefly discuss the

606

A. Peplinski et al.

Time (seconds)

10

1

0.1

(a)

(b)

solver time linear scaling

16

32

64

128

256

Node number

Fig. 1 (a) Volume visualisation of that part of the domain covered by refinement levels higher than one for the turbulent flow around a wing profile. The wing vicinity and wake region are resolved and a colour indicates different refinement levels. (b) Strong scaling of the non-conforming Nek5000 solver for the same case performed on Beskow. The plot shows the time per time step as a function of node number. Each node consists of 32 cores

strong scaling results. We omit here a weak scaling test, as Nek5000 uses iterative solvers and with the current example we cannot provide meaningful data. The initial coarse and conforming mesh consisted of 2190 elements with polynomial order N = 7 and was evolved for 7.2 time units c/U∞ to evolve the refinement process using spectral error indicators [13, 14], and allowing for 6 refinements levels. The resulting non-conforming grid was built of 224,272 elements with 76.37 × 106 degrees of freedom, resolving the wing surface and the wake, Fig. 1a. This final mesh was used to test the parallel performance of the non-conforming solver using the petascale Cray XC40 system Beskow at PDC (Stockholm). This system consists of 2060 nodes with 32 cores per node and 2.438 PFlops peak performance. We compare our results with the scaling tests of the conforming Nek5000 presented in Offermans et al. [15]. The most relevant test in this article is pipe flow at Reτ = 360 (upper-right plot in their Fig. 5), as it is similar in size with the discussed wing case. We should mention here that our goal is not to improve the parallel performance of the conforming code, but rather to retain it despite of a work imbalance introduced by an additional operator in the direct stiffness summation of the non-conforming solver. To be able to compare to the conforming solver, we focus on the time evolution loop only, excluding code initialisation, finalisation, mesh rebuilding within AMR and I/O operations. The result of the strong scaling test is presented in Fig. 1b showing the time per time step as a function of node count. This plot is almost identical with the reference one in [15]. Both show slight super-linear scaling between 32 and 256 nodes despite growing work imbalance for the non-conforming solver. We also reach the strong scaling limit at around 256 nodes, which for the conforming solver on Beskow was estimated to be between 30,000 and 50,000 degrees of freedom per core [15]. This shows that the parallel performance of the non-conforming and conforming solvers is almost the same and proves the efficiency of our implementation.

Non-conforming Elements in Nek5000

607

The maximum number of the compute nodes used in the test was not set by the parallel properties of the non-conforming Nek5000, but by the quality of the domain partitioning provided by ParMETIS. Within ExaFLOW we developed a new grid partitioning scheme for Nek5000 (not discussed in this paper) that takes into account a core distribution among the nodes, and consists of two steps: inter- and intra-node partitioning. Although this two-level partitioning scheme significantly improves the efficiency of a coarse grid operations for XXT, especially during the setup phase, it relies on the quality of an inter-node partitioning. If the first step gives subdomains with disjoint graphs, the second step cannot be performed. We found that the probability of getting disjoint graphs increases with decreasing number of elements per node, virtually prohibiting the runs with less than 1000 elements per node. However, this limit can differ between simulations. We note however that in the standard production use of the solver this limitation is not critical, as according to [15] it is usually close to the strong scaling limit of conforming Nek5000.

6 Conclusions Within the ExaFLOW project we developed a fully functional SEM-based h-type adaptive mesh refinement (AMR) solver for the incompressible Navier–Stokes equations. This allows for much larger flow cases to be run at reduced cost, as the high resolution grid is placed only in those region where it is needed. At the same time the simulation quality is improved, as the computational error can be controlled during the run. We have optimised for non-conforming meshes the pressure preconditioners based on the additive overlapping Schwarz and hybrid Schwarz-multigrid methods. To achieve this we modified the base functions for the assembly of a coarse-grid operator to remove hanging nodes, and redefined the direct stiffness summation operator to include spectral interpolation at the non-conforming faces and edges. We introduced two operators JL QQT JTL and JL QQT J−1 L for the different steps in the pressure calculation. The last crucial modification was the diagonalisation of the global mass matrix. Using real flow cases we show our AMR implementation to be correct and efficient. An important success is the fact that parallel performance of the conforming and non-conforming solvers is very similar, despite the increased complexity of the non-conforming one. In the future we are going to investigate other definitions of the weight matrix for the hybrid Schwarz-multigrid method, and to test different pressure preconditioners based on the restricted additive Schwarz method [2, 3]. We are going as well to work on the quality of the graph partition, as the two-level partitioning would not accept disjoint graphs on the node’s subdomain.

608

A. Peplinski et al.

Acknowledgements We would like to thank Niclas Jansson for sharing his expertise on adaptive mesh refinement. The work presented in this publication was supported by a European Commission Horizon 2020 project grant entitled “ExaFLOW: Enabling Exascale Fluid Dynamics Simulation” (grant reference 671571). Computer time was provided by Swedish National Infrastructure for Computing (SNIC) at PDC (KTH Stockholm) and by ExaFLOW at HLRS Stuttgart.

References 1. Burstedde, C., Wilcox, L., Ghattas, O.: p4est: scalable algorithms for parallel adaptive mesh refinement on forests of octrees. SIAM J. Sci. Comput. 33(3), 1103–1133 (2011) 2. Cai, X.C., Sarkis, M.: A restricted additive Schwarz preconditioner for general sparse linear systems. SIAM J. Sci. Comput. 21(2), 792–797 (1999) 3. Efstathiou, E., Gander, M.J.: Why restricted additive Schwarz converges faster than additive Schwarz. BIT Numer. Math. 43(5), 945–959 (2003) 4. Fischer, P.F.: An overlapping Schwarz method for spectral element solution of the incompressible Navier–Stokes equations. J. Comput. Phys. 133, 84–101 (1997) 5. Fischer, P.F., Lottes, J.W.: Hybrid Schwarz-Multigrid Methods for the Spectral Element Method: Extensions to Navier-Stokes, pp. 35–49. Springer, Berlin (2005) 6. Fischer, P., Miller, N., Tufo, H.: An overlapping Schwarz method for spectral element simulation of three-dimensional incompressible flow. In: Parallel Solution of Partial Differential Equations, pp. 159–180. Springer, New York (2000) 7. Fischer, P.F., Kruse, G.W., Loth, F.: Spectral element methods for transitional flows in complex geometries. J. Sci. Comput. 17(1–4), 81–98 (2002) 8. Fischer, P.F., Lottes, J.W., Kerkemeier, S.G.: Nek5000 Web page (2008). http://nek5000.mcs. anl.gov 9. Fryxell, B., Olson, K., Ricker, P., Timmes, F.X., Zingale, M., Lamb, D.Q., MacNeice, P., Rosner, R., Truran, J.W., Tufo, H.: FLASH: an adaptive mesh hydrodynamics code for modeling astrophysical thermonuclear flashes. Astrophys. J. Suppl. Ser. 131, 273–334 (2000) 10. Karypis, G., Kumar, V.: A fast and high quality multilevel scheme for partitioning irregular graphs. SIAM J. Sci. Comput. 20(1), 359–392 (1998) 11. Kruse, G.W.: Parallel nonconforming spectral element solution of the incompressible Navier– Stokes equations in three dimensions. Ph.D. Thesis, Brown University, Providence, 1997. UMI Order No. GAX97-38573 12. Lottes, J.W., Fischer, P.F.: Hybrid multigrid/Schwarz algorithms for the spectral element method. J. Sci. Comput. 24(1), 45–78 (2005) 13. Mavriplis, C.: A posteriori error estimators for adaptive spectral element techniques. In: Wesseling, P. (ed.) Proceedings of the Eighth GAMM-Conference on Numerical Methods in Fluid Mechanics. Notes on Numerical Fluid Mechanics, pp. 333–342 (1990) 14. Offermans, N.: Towards adaptive mesh refinement in Nek5000. Licentiate Thesis, KTH, Mechanics, 2017 15. Offermans, N., Marin, O., Schanen, M., Gong, J., Fischer, P., Schlatter, P.: On the strong scaling of the spectral element solver Nek5000 on petascale systems. In: Proceedings of the Exascale Applications and Software Conference 2016, Stockholm, April 26–29, 2016 (2016) 16. Patera, A.T.: A spectral element method for fluid dynamics: Laminar flow in a channel expansion. J. Comput. Phys. 54(3), 468–488 (1984) 17. Peplinski, A., Fischer, P.F., Schlatter, P.: Parallel performance of h-type adaptive mesh refinement for Nek5000. In: Proceedings of the Exascale Applications and Software Conference 2016, Stockholm, April 26–29, 2016, pp. 4:1–4:9 (2016)

Non-conforming Elements in Nek5000

609

18. Rosenberg, D., Fournier, A., Fischer, P., Pouquet, A.: Geophysical-astrophysical spectralelement adaptive refinement (GASpAR): object-oriented h-adaptive fluid dynamics simulation. J. Comput. Phys. 215(1), 59–80 (2006) 19. Vinuesa, R., Negi, P.S., Atzori, M., Hanifi, A., Henningson, D.S., Schlatter, P.: Turbulent boundary layers around wing sections up to Rec = 1,000,000. Int. J. Heat Fluid Flow 72, 86–99 (2018)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Sparse Approximation of Multivariate Functions from Small Datasets Via Weighted Orthogonal Matching Pursuit Ben Adcock and Simone Brugiapaglia

1 Introduction In recent years, a new class of approximation strategies based on compressive sensing (CS) has been shown to be able to substantially lessen the curse of dimensionality in the context of approximation of multivariate functions from pointwise data, with applications to the uncertainty quantification of partial differential equations with random inputs. Based on random sampling from orthogonal polynomial systems and on weighted 1 minimization, these techniques are able to accurately recover a sparse approximation to a function of interest from a smallsized datasets of pointwise samples. In this paper, we show the potential of weighted greedy techniques as an alternative to convex minimization programs based on weighted 1 minimization in this context. The contribution of this paper is twofold. First, we propose a weighted orthogonal matching pursuit (WOMP) algorithm based on a rigorous derivation of the corresponding greedy index selection strategy. Second, we numerically show that WOMP is a promising alternative to convex recovery programs based on weighted 1 minimization, thanks to its ability to compute sparse approximations with an accuracy comparable to those computed via weighted 1 minimization, but with a considerably lower computational cost when the target sparsity level (and, hence, the number of WOMP iterations) is small enough. It is also worth observing here that WOMP computes approximations that are exactly sparse, as opposed to approaches based on weighted 1 minimization, which provide compressible approximations in general.

B. Adcock · S. Brugiapaglia () Simon Fraser University, Burnaby, BC, Canada e-mail: [email protected]; [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_49

611

612

B. Adcock and S. Brugiapaglia

Brief Literature Review Various approaches for multivariate function approximation based on CS with applications to uncertainty quantification can be found in [1, 3–6, 11–13, 17]. An overview of greedy methods for sparse recovery in CS and, in particular of OMP, can be found in [7, Chapter 3.2]. For a general review on greedy algorithms, we refer the reader to [15] and references therein. Some numerical experiments on a weighted variant of OMP have been performed in the context of CS methods for uncertainty quantification in [4]. Weighted variants of OMP have also been considered in [10, 16], but the weighted procedure is tailored for specific signal processing applications and the term “weighted” does not refer to the weighted sparsity setting of [14] employed here. To the authors’ knowledge, the weighted variant of OMP considered in this paper seems to have been proposed here for the first time. Organization of the Paper In Sect. 2 we describe the setting of sparse multivariate function approximation in orthonormal systems via random sampling and weighted 1 minimization. Then, in Sect. 3 we formally derive a strategy for the greedy selection in the weighted sparsity setting and present the WOMP algorithm. Finally, we numerically show the effectiveness of the proposed technique in Sect. 4 and give our conclusions in Sect. 5.

2 Sparse Multivariate Function Approximation We start by briefly introducing the framework of sparse multivariate function approximation from pointwise samples and refer the reader to [3] for further details. Our aim is to approximate a function defined over a high-dimensional domain f : D → C,

with D = (−1, 1)d ,

where d ' 1, from a dataset of pointwise samples f (t1 ), . . . , f (tm ). Let ν be a probability measure on D and let {φj }j ∈Nd be an orthonormal basis for the Hilbert 0

space L2ν (D). In this paper, we will consider {φj }j ∈Nd to be a tensorized family of 0 Legendre or Chebyshev orthogonal polynomials, with ν being the uniform or the Chebyshev measure on D, respectively. Assuming that f ∈ L2ν (D) ∩ L∞ (D), we consider the series expansion  x j φj . f = j ∈Nd0

Then, we choose a finite set of multi-indices * ⊆ Nd0 with |*| = N and obtain the truncated series expansion f* =

 j ∈*

x j φj .

Sparse Approximation of Multivariate Functions from Small Datasets Via. . .

613

In practice, a convenient choice for * is the hyperbolic cross of order s, i.e. * :=

⎧ ⎨ ⎩

d @

j ∈ Nd0 :

(jk + 1) ≤ s

k=1

⎫ ⎬ ⎭

,

due to the moderate growth of N with respect to d. Now, assuming we collect m N pointwise samples independently distributed according to ν, namely, f (t1 ), . . . , f (tm ),

with

i.i.d.

t1 , . . . , tm ∼ ν,

the approximation problem can be recasted as a linear system Ax* = y + e,

(1)

with x* = (xj )j ∈* ∈ CN , and where the sensing matrix A ∈ Cm×N and the measurement vector y ∈ Cm are defined as 1 Aij := √ φj (ti ), m

1 yi := √ f (ti ), m

∀i ∈ [m], ∀j ∈ [N],

(2)

with [k] := {1, . . . , k} for every k ∈ N. The vector e ∈ Cm accounts for the truncation error introduced by * and satisfies e2 ≤ η, where η > 0 is an a priori upper bound to the truncation L∞ (D)-error, namely f − f* L∞ (D) ≤ η. A sparse approximation to the vector can be then computed by means of weighted 1 minimization. Given weights w ∈ RN with w > 0 (where the inequality is read componentwise), < recall that the weighted 1 norm of a vector z ∈ CN is defined as z1,w := j ∈[N] |zj |wj . We can compute an approximation xˆ* to x* by solving the weighted quadratically-constrained basis pursuit (WQCBP) program xˆ* ∈ arg min z1,w ,

s.t.

z∈CN

Az − y2 ≤ η,

(3)

where the weights w ∈ RN are defined as wj = φj L∞ (D) .

(4)

The effectiveness of this particular choice of w is supported by theoretical results and it has been validated from the numerical viewpoint (see [1, 3]). The resulting approximation fˆ* to f is finally defined as fˆ* :=



(xˆ* )j φj .

j ∈*

614

B. Adcock and S. Brugiapaglia

In this setting, stable and robust recovery guarantees in high probability can be shown for the approximation errors f − f* L2ν (D) and f − f* L∞ under a ν (D) sufficient condition on the number of samples of the form m  s γ · polylog(s, d), with γ = 2 or γ = log(3)/ log(2) for tensorized Legendre or Chebyshev polynomials, respectively, hence lessening the curse of dimensionality to a substantial extent (see [3] and references therein). We also note in passing that decoders such as the weighted LASSO or the weighted square-root LASSO can be considered as alternatives to (3) for weighted 1 minimization (see [2]).

3 Weighted Orthogonal Matching Pursuit In this paper, we consider greedy sparse recovery strategies to find sparse approximate solutions to (1), as alternatives to the WQCBP optimization program (3). With this aim, we propose a variation of the OMP algorithm to the weighted setting. Before introducing weighted OMP (WOMP) in Algorithm 1, let us recall the rationale behind the greedy index selection rule of OMP (corresponding to Algorithm 1 with λ = 0 and w = 1). For a detailed introduction to OMP, we refer the reader to [7, Section 3.2]. Given a support set S ⊆ [N], OMP solves the least-squares problem min G0 (z) s.t. supp(z) ⊆ S,

z∈CN

where G0 (z) := y − Az22 . In OMP, the support S is iteratively enlarged by one index at the time. Namely, we consider the update S ∪ {j }, where the index j ∈ [N] is selected in a greedy fashion. In particular, assuming that A has 2 -normalized columns, it is possible to show that (see [7, Lemma 3.3]) min G0 (x + tej ) = G0 (x) − |(A∗ (y − Ax))j |2 . t ∈C

(5)

This leads to the greedy index selection rule operated by OMP, which prescribes the selection of an index j ∈ [N] that maximizes the quantity |(A∗ (y − Ax))j |2 . We will use this simple intuition to extend OMP to the weighted case by replacing the function G0 with a suitable function Gλ that takes into account the data-fidelity term and the weighted sparsity prior at the same time. Let us recall that, given a set of weights w ∈ RN with w > 0, the weighted 0 norm of a vector z ∈ CN is defined as the quantity (see [14])1 z0,w :=



wj2 .

j ∈supp(z)

1 The

term “norm” here is an abuse of language, but we will stick to it due to its popularity.

Sparse Approximation of Multivariate Functions from Small Datasets Via. . .

615

Notice that when w = 1, then  · 0,w =  · 0 is the standard 0 norm. Given λ ≥ 0, we define the function Gλ (z) := y − Az22 + λz0,w .

(6)

The tradeoff between the data-fidelity constraint and the weighted sparsity prior is balanced via the choice of the regularization parameter λ. Applying the same rationale employed in OMP for the greedy index selection and replacing G0 with Gλ leads to Algorithm 1, which corresponds to OMP when λ = 0 and w = 1. Algorithm 1 Weighted orthogonal matching pursuit (WOMP) Inputs: • • • • •

A ∈ Cm×N : sampling matrix, with 2 -normalized columns; y ∈ Cm : vector of samples; w ∈ RN : weights; λ ≥ 0: regularization parameter; K ∈ N: number of iterations.

Procedure: 1. Let xˆ0 = 0 and S0 = ∅; 2. For k = 1, . . . , K: a. Find jk ∈ arg max λ (xk−1 , Sk−1 , j ), with λ as in (7); j ∈[N]

b. Define Sk = Sk−1 ∪ {jk }; c. Compute xˆk ∈ arg min Av − y2 s.t. supp(v) ⊆ Sk . v∈CN

Output: • xˆK ∈ CN : approximate solution to Az = y.

Remark 1 The 2 -normalization of the columns of A is a necessary condition to apply Algorithm 1. If A does not satisfy this hypothesis, is suffices to apply WOMP 3 = y, where A 3 = AM −1 and M is the matrix containing to the normalized system Az the 2 norms of the columns of A on the main diagonal and zeroes elsewhere. The 3 = y computed via WOMP is then rescaled as M xˆK , approximate solution xˆK to Az which approximately solves Az = y. The following proposition justifies the weighted variant of OMP considered in Algorithm 1. In order to minimize Gλ as much as possible, at each iteration, WOMP selects the index j that maximizes the quantity λ (x, S, j ) defined in (7). The following proposition makes the role of the quantity λ (x, S, j ) transparent, generalizing relation (5) to the weighted case, under suitable conditions on A and x that are verified at each iteration of Algorithm 1.

616

B. Adcock and S. Brugiapaglia

Proposition 1 Let λ ≥ 0, S ⊆ [N], A ∈ Cm×N with 2 -normalized columns, and x ∈ CN satisfying x ∈ arg min y − Az2 s.t. supp(z) ⊆ S. z∈CN

Then, for every j ∈ [N], the following holds: min Gλ (x + tej ) = Gλ (x) − λ (x, S, j ), t ∈C

where Gλ is defined as in (6), λ : CN × 2[N] × [N] → R is defined by ⎧   ∗ (y − Ax)) |2 − λw 2 , 0 ⎪ ⎪ max |(A if j ∈ /S j ⎪ j ⎨   2

λ (x, S, j ) := max λwj − |xj |2 , 0 if j ∈ S and xj =  0 ⎪ ⎪ ⎪ ⎩0 if j ∈ S and xj = 0. (7) Proof Throughout the proof, we will denote the residual as r := y − Ax. Let us first assume j ∈ / S. In this case, we compute Gλ (x + tej ) = y − A(x + tej )22 + λx + tej 0,w = r22 + |t|2 − 2 Re(t¯(A∗ r)j ) + λ(1 − δt,0 )wj2 +λx0,w , F GH I =:h(t )

where δx,y is the Kronecker delta function. In particular, we have h(t) =

⎧ ⎨0

if t = 0

⎩|t|2 − 2 Re(t¯(A∗ r)j ) + λw2 j

if t ∈ C \ {0}.

Now, if (A∗ r)j = 0, then h(t) is minimized for t = 0 and mint ∈C G(x + tej ) = G(x). On the other hand, if (A∗ r)j = 0, by arguing similarly to [7, Lemma 3.3], we see that min h(t) = −|(A∗ r)j |2 + λwj2 ,

t ∈C\{0}

where the minimum is realized for some t ∈ C with |t| = |(A∗ r)j | = 0. In summary,     min h(t) = min −|(A∗ r)j |2 + λwj2 , 0 = − max |(A∗ r)j |2 − λwj2 , 0 , t ∈C

which concludes the case j ∈ / S.

Sparse Approximation of Multivariate Functions from Small Datasets Via. . .

617

Now, assume j ∈ S. Since the vector xS = x|S ∈ C|S| is a least-squares solution to AS z = y, it satisfies A∗S (y − AS xS ) = 0 and, in particular, (A∗ r)j = 0. (Here, AS ∈ Cm×|S| denotes the submatrix of A corresponding to the columns in S). Therefore, arguing similarly as before, we have G(x + tej ) = r22 + |t|2 + λ(1 − δt,−xj )wj2 +λx − xj ej 0,w . F GH I =:(t )

Considering only the terms depending on t, it is not difficult to see that min (t) = min{|xj |2 , λwj2 }. t ∈C

As a consequence, for every j ∈ S, we obtain min G(x + tej ) = r22 + λx − xj ej 0,w + min{|xj |2 , λwj2 } t ∈C

= G(x) + min{|xj |2 , λwj2 } − λ(1 − δxj ,0 )wj2 . The results above combined with simple algebraic manipulations lead to the desired result. & %

4 Numerical Results In this section, we show the effectiveness of WOMP (Algorithm 1) in the sparse multivariate function approximation setting described in Sect. 2. In particular, we choose the weights w as in (4). We consider the function ⎛ f (t) = ln ⎝d + 1 +

d 

⎞ tk ⎠ ,

with d = 10.

(8)

k=1

We let {φj }j ∈Nd be the Legendre and Chebyshev bases and ν be the respective 0

orthogonality measure. In Figs. 1 and 2 we show the relative L2ν (D)-error of the approximate solution xˆK computed via WOMP as a function of iteration K, for different values of the regularization parameter λ in order to solve the linear system Az = y, where A and y are defined by (2) and where the 2 -normalization of the columns of A is taken into account according to Remark 1. We consider λ = 0 (corresponding to OMP) and λ = 10−k , with k = 3, 3.5, 4, 4.5, 5. Here, * is the hyperbolic cross of order s = 10, corresponding to N = |*| = 571. Moreover, we consider m = 60 and m = 80. The results are averaged over 25 runs and the L2ν (D)error is computed with respect to a reference solution approximated via least squares

618

B. Adcock and S. Brugiapaglia

Fig. 1 Plot of the mean relative L2ν (D)-error as a function of the number of iterations K of WOMP (Algorithm 1) for different values of the regularization parameter λ for the approximation of the function f defined in (8) and using Legendre polynomials. The accuracy of WOMP is compared with those of QCBP and WQCBP

Fig. 2 The same experiment as in Fig. 1, with Chebyshev polynomials

and using 20N = 11,420 random i.i.d. samples according to ν. We compare the WOMP accuracy with the accuracy obtained via the QCBP program (3) with η = 0 and WQCBP with tolerance parameter η = 10−8 . To solve these two programs we use CVX Version 1.2, a package for specifying and solving convex programs [8, 9]. In CVX, we use the solver ‘mosek’ and we set CVX precision to ‘high’. Figures 1 and 2 show the benefits of using weights as compared to the unweighted OMP approach, when the parameter λ is tuned appropriately. A good choice of λ for the setting considered here seem to be between 10−4.5 and 10−3.5 . We also observe that WOMP is able to reach similar level of accuracy as WQCBP. An interesting feature of WOMP with respect to OMP is its better stability. We observe than after the m-th iteration, the OMP accuracy starts getting substantially worse. This can be explained by the fact that when K approaches N, OMP tends to destroy sparsity by fitting the data too much. This phenomenon is not observed in WOMP, thanks to its

Sparse Approximation of Multivariate Functions from Small Datasets Via. . .

619

Fig. 3 Plot of the support size of xˆK as a function of the number of iterations K for WOMP in the same setting as in Figs. 1 and 2, with Legendre (left) and Chebyshev (right) polynomials. The larger the regularization parameter λ, the sparser solution (in the left plot, the curves relative to λ = 10−4.5 and λ = 10−4 overlap. In the right plot, the same happens for λ = 10−4 and λ = 10−3.5 ) Table 1 Comparison of the computing times for WQCBP and K = 25 iterations of WOMP Basis

m QCBP

WQCBP OMP

WOMP with λ as below 10−5 10−4.5 10−4

10−3.5

10−3

Legendre Legendre Chebyshev Chebyshev

60 80 60 80

2.0e−01 2.1e−01 1.9e−01 2.1e−01

1.3e−02 1.5e−02 1.3e−02 1.5e−02

1.2e−02 1.4e−02 1.2e−02 1.4e−02

1.2e−02 1.3e−02 1.2e−02 1.4e−02

1.9e−01 2.1e−01 1.9e−01 2.1e−01

1.6e−02 1.7e−02 1.5e−02 1.7e−02

1.2e−02 1.3e−02 1.2e−02 1.3e−02

1.3e-02 1.4e−02 1.2e−02 1.4e−02

ability to keep the support of xˆ k small via the explicit enforcement of the weighted sparsity prior (see Fig. 3). We show the better computational efficiency of WOMP with respect to the convex minimization programs QCBP and WQCBP solved via CVX by tracking the runtimes for the different approaches. In Table 1 we show the running times for the different recovery strategies. The running times for WOMP are referred to K = 25 iterations, sufficient to reach the best accuracy for every value of λ as shown in Figs. 1 and 2. Moreover, the computational times for WOMP take into account the 2 -normalization of the columns of A (see Remark 1). WOMP consistently outperforms convex minimization, being more than ten times faster in all cases. We note that in this comparison a key role is played by the parameter K or, equivalently, by the sparsity of the solution. Indeed, in this case, considering a larger value of K would result is a slower performance of WOMP, but it would not improve the accuracy of the WOMP solution (see Figs. 1 and 2).

620

B. Adcock and S. Brugiapaglia

5 Conclusions We have considered a greedy recovery strategy for high-dimensional function approximation from a small set of pointwise samples. In particular, we have proposed a generalization of the OMP algorithm to the setting of weighted sparsity (Algorithm 1). The corresponding greedy selection strategy is derived in Proposition 1. Numerical experiments show that WOMP is an effective strategy for highdimensional approximation, able to reach the same accuracy level of WQCBP while being considerably faster when the target sparsity level is small enough. A key role is played by the regularization parameter λ, which may be difficult to tune due to its sensitivity to the parameters of the problem (m, s, and d), and on the polynomial basis employed. In other applications, where explicit formulas for the weights as (4) are not available, there might also be a nontrivial interplay between λ and w. In summary, despite the promising nature of the numerical experiments illustrated in this paper, a more extensive numerical investigation is needed in order to study the sensitivity of WOMP with respect to λ. Moreover, a theoretical analysis of the WOMP approach might highlight practical recipe for the choice of this parameter, similarly to [2]. This type of analysis may also help identifying the sparsity regime where WOMP outperforms weighted 1 minimization, which, in turn, could be formulated in terms of suitable assumptions on the regularity of f . These questions are beyond the scope of this paper and will be object of future work. Acknowledgements The authors acknowledge the support of the Natural Sciences and Engineering Research Council of Canada through grant number 611675, and of the Pacific Institute for the Mathematical Sciences (PIMS) Collaborative Research Group “High-Dimensional Data Analysis”. S.B. also acknowledges the support of the PIMS Postdoctoral Training Centre in Stochastics.

References 1. Adcock, B.: Infinite-dimensional compressed sensing and function interpolation. Found. Comput. Math. 18(3), 661–701 (2018) 2. Adcock, B., Bao, A., Brugiapaglia, S.: Correcting for unknown errors in sparse highdimensional function approximation (2017). Preprint. arXiv:1711.07622 3. Adcock, B., Brugiapaglia, S., Webster, C.G.: Compressed sensing approaches for polynomial approximation of high-dimensional functions. In: Boche, H., Caire, G., Calderbank, R., März, M., Kutyniok, G., Mathar R. (eds.) Compressed Sensing and Its Applications: Second International MATHEON Conference 2015, pp. 93–124. Springer International Publishing, Cham (2017) 4. Bouchot, J.-L., Rauhut, H., Schwab C.: Multi-level compressed sensing Petrov-Galerkin discretization of high-dimensional parametric PDEs (2017). Preprint. arXiv:1701.01671 5. Chkifa, A., Dexter, N., Tran, H., Webster, C.G.: Polynomial approximation via compressed sensing of high-dimensional functions on lower sets. Math. Comp. 87(311), 1415–1450 (2018) 6. Doostan, A., Owhadi, H.: A non-adapted sparse approximation of PDEs with stochastic inputs. J. Comput. Phys. 230(8), 3015–3034 (2011)

Sparse Approximation of Multivariate Functions from Small Datasets Via. . .

621

7. Foucart, S., Rauhut, H.: A Mathematical Introduction to Compressive Sensing. Birkhäuser Basel (2013) 8. Grant, M., Boyd, S.: Graph implementations for nonsmooth convex programs. In: Blondel, V., Boyd, S., Kimura, H. (eds.) Recent Advances in Learning and Control. Lecture Notes in Control and Information Sciences, pp. 95–110. Springer, Berlin (2008) 9. Grant, M., Boyd, S.: CVX: Matlab software for disciplined convex programming, version 2.1. http://cvxr.com/cvx (2014) 10. Li, G.Z., Wang, D.Q., Zhang, Z.K., Li, Z.Y.: A weighted OMP algorithm for compressive UWB channel estimation. In: Applied Mechanics and Materials, vol. 392, pp. 852–856. Trans Tech Publications, Zurich (2013) 11. Mathelin, L., Gallivan, K.A.: A compressed sensing approach for partial differential equations with random input data. Commun. Comput. Phys. 12(4), 919–954 (2012) 12. Peng, J., Hampton, J., Doostan, A.: A weighted 1 -minimization approach for sparse polynomial chaos expansions. J. Comput. Phys. 267, 92–111 (2014) 13. Rauhut, H., Schwab, C.: Compressive sensing Petrov-Galerkin approximation of highdimensional parametric operator equations. Math. Comp. 86(304), 661–700 (2017) 14. Rauhut, H., Ward, R.: Interpolation via weighted 1 minimization. Appl. Comput. Harmon. Anal. 40(2), 321–351 (2016) 15. Temlyakov, V.N.: Greedy approximation. Acta Numer. 17, 235–409 (2008) 16. Xiao-chuan, W., Wei-bo, D., Ying-ning, D.: A weighted OMP algorithm for Doppler superresolution. In: 2013 Proceedings of the International Symposium on Antennas & Propagation (ISAP), vol. 2, pp. 1064–1067. IEEE, Piscataway (2013) 17. Yang, X., Karniadakis, G.E.: Reweighted 1 minimization method for stochastic elliptic differential equations. J. Comput. Phys. 248, 87–108 (2013)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

On the Convergence Rate of Hermite-Fejér Interpolation Shuhuang Xiang and Guo He

1 Introduction For an arbitrarily given system of points {x1 , x2 , . . . , xn(n) }∞ n=1 , (n)

(n)

(1)

Faber [3] in 1914 showed that there exists a continuous function f (x) in [−1, 1] for which the Lagrange interpolation sequence Ln [f ] (n = 1, 2, . . .) is not uniformly convergent to f in [−1, 1], where ωn (x) = (x − x1(n) )(x − x2(n) ) · · · (x − xn(n) ) Ln [f ](x) =

n 

(n) f (xk(n) )(n) k (x), k (x) =

k=1

ωn (x) . (n) (n)  ωn (xk )(x − xk )

(2)

Whereas, based on the Chebyshev pointsystem  (n)

xk

= cos

2k − 1 π , 2n

k = 1, 2, . . . , n,

n = 1, 2, . . . ,

(3)

S. Xiang Department of Mathematics, Central South University, Changsha, China e-mail: [email protected] G. He () Department of Mathematics, Jinan University, Guangzhou, Guangdong, China © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_50

623

624

10

S. Xiang and G. He f(x)=sin(x)

0

10-5

||f(x)-H

2n-1

(f,x)|| 10

||f(x)-Ln(f,x)|| 10

-10

f(x)=1/(1+25x2)

100

-5

10-5

||f(x)-H*2n-1(f,x)||

10-10

||f(x)-H

10-15 10-20 100

f(x)=|x|3

100

2n-1

||f(x)-H

(f,x)||

n ||f(x)-H* (f,x)|| 2n-1

10-15 101

102

103

100

101

(f,x)||

102

||f(x)-H*2n-1(f,x)||

10-10 100

103

101

n=1:10:1000

n=1:10:1000

2n-1

||f(x)-Ln(f,x)||

||f(x)-L (f,x)||

102

103

n=1:10:1000

∗ Fig. 1 H2n−1 (f, x) − f (x)∞ , Ln (f, x) − f (x)∞ and H2n−1 (f, x) − f (x)∞ at x = −1 : 1 3 0.001 : 1 by using Chebyshev pointsystem (3) for f (x) = sin(x), f (x) = 1+25x 2 and f (x) = |x| , respectively

Fejér [4] in 1916 proved that if f ∈ C[−1, 1], then there is a unique polynomial H2n−1 (f, x) of degree at most 2n − 1 such that limn→∞ H2n−1 (f ) − f ∞ = 0, where H2n−1 (f, x) is determined by (n)

 H2n−1 (f, xk ) = 0,

(n)

(n)

H2n−1 (f, xk ) = f (xk ),

k = 1, 2, . . . , n.

(4)

This polynomial is known as the Hermite-Fejér interpolation polynomial. It is of particular notice that the above Hermite-Fejér interpolation polynomial converges much slower compared with the corresponding Lagrange interpolation polynomial at the Chebyshev pointsystem (3) (see Fig. 1). To get fast convergence, the following Hermite-Fejér interpolation of f (x) at nodes (1) is considered [6, 7]: ∗ H2n−1 (f, x) =

n 

(n)

(n)

f (xk )hk (x) +

k=1

n 

f  (xk )bk (x), (n)

(n)

(5)

k=1

 2  2 (n) (n) (n) (n) (n) (n) (n) where hk (x) = vk (x) k (x) , bk (x) = (x − xk ) k (x) and vk (x) = 1 − (x − xk(n) )

ωn (xk ) (n)

(n) ωn (xk )

.

Fejér [5] and Grünwald [7] also showed that the convergence of the HermiteFejér interpolation of f (x) also depends on the choice of the nodes. The pointsystem (1) is called normal if for all n (n)

vk (x) ≥ 0,

k = 1, 2, . . . , n,

x ∈ [−1, 1],

(6)

while the pointsystem (1) is called strongly normal if for all n vk(n) (x) ≥ c > 0, for some positive constant c.

k = 1, 2, . . . , n,

x ∈ [−1, 1]

(7)

On the Convergence Rate of Hermite-Fejér Interpolation

625

Fejér [5] (also see Szegö [12, pp 339]) showed that for the zeros of Jacobi (α,β) (x) of degree n (α > −1, β > −1) polynomial Pn (n)

vk (x) ≥ min{−α, −β}

for − 1 < α ≤ 0, −1 < β ≤ 0, k = 1, 2, . . . , n and x ∈ [−1, 1].

For (strongly) normal pointsystems, Grünwald [7] showed that for every f ∈ ∗ C 1 (−1, 1), limn→∞ H2n−1 (f ) − f ∞ = 0 if {xk(n) } is strongly normal satisfying (7) and {f  (xk )} satisfies (n)

|f  (xk )| < nc−δ (n)

for some given positive number δ,

k = 1, 2, . . . ,

n = 1, 2, . . . ,

∗ while limn→∞ H2n−1 (f ) − f ∞ = 0 in [−1 + #, 1 − #] for each fixed 0 < # < 1

if {xk } is normal and {f  (xk )} is uniformly bounded for n = 1, 2, . . ..1 Moreover, Szabados [11] showed the convergence of the Hermite-Fejér interpolation (5) at the Chebyshev pointsystem (3) satisfies (n)

(n)

∗ f − H2n−1 (f )∞ = O(1)f − p∗ C 1 [−1,1]

(8)

where p∗ is the best approximation polynomial of f with degree at most 2n − 1 and f − p∗ C 1 [−1,1] = max0≤j ≤1 f (j ) − p∗ (j ) ∞ . Hermite-Fejér interpolation has plenty of use in computer geometry aided geometric design with boundary conditions including derivative information. The convergence rate under the infinity norm has been extensively studied in [5– 7, 11, 14]. The efficient algorithm on the fast implementation of Hermite-Fejér interpolation at zeros of Jacobi polynomial can be found in [17]. In this paper, the following convergence rates of Hermite-Fejér interpolation ∗ H2n−1 (f, x) at Gauss-Jacobi pointsystems are considered. • If f is analytic in Eρ with |f (z)| ≤ M, then

∗ f (x) − H2n−1 (f, x)∞

⎧ # $ ⎪ ⎪ 4τn M[2nρ 2 + (1 − 2n)ρ] ⎪ ⎪ , γ ≤ 0, ⎪ ⎨O (ρ − 1)2 ρ 2n # $ , γ = max{α, β} = 2+2γ [2nρ 2 + (1 − 2n)ρ] ⎪ ⎪ ⎪O n , γ > 0 ⎪ ⎪ ⎩ (ρ − 1)2 ρ 2n

(9)

1 In

fact, Grünwald in [7] considered more general cases with any vector {dk(n) } instead of

{f  (xk(n) )}.

626

S. Xiang and G. He

where ⎧ −1.5−min{α,β} log n), if − 1 < min{α, β} ≤ γ ≤ − 1 ⎪ ⎨ O(n 2 1 τn = O(n2γ −min{α,β}− 2 ), if − 1 < min{α, β} ≤ − 12 < γ ≤ 0 . ⎪ ⎩ O(n2γ ), if − 12 < min{α, β} ≤ γ

(10)

• If f (x) has an absolutely continuous (r − 1)st derivative f (r−1) on [−1, 1] for an integer r ≥ 3, and a rth derivative f (r) of bounded variation Vr = Var(f (r) ) < ∞, then ⎧   ⎪ ⎨ O n−r log n , γ ≤ − 1 , 2 ∗   (11) (f, x)∞ = f (x) − H2n−1 2γ −r+1 ⎪ , γ > − 12 , ⎩O n while if f (x) is differentiable and f  (x) is bounded on [−1, 1], then ∗ f (x) − H2n−1 (f, x)∞

⎧   ⎪ ⎨ O n−1 log n , γ ≤ − 1 , 2   = 2γ ⎪ γ > − 12 . ⎩O n ,

Comparing these results with f (x) − H2n−1 (f, x) =

⎧ ⎨

  O n−1 log n , if γ ≤ − 12

⎩ O(n2γ ),

if γ > − 12

, (Vértesi [14]),

∗ which is sharp and attainable (see Fig. 2), we see that H2n−1 (f, x) converges much faster than H2n−1 (f, x) for analytic functions or functions of higher regularities (see Fig. 1). Particularly, H2n−1 (f, x) diverges at Gauss-Jacobi pointsystems with γ ≥ 0, ∗ whereas, H2n−1 (f, x) converges for functions analytic in the Bernstein ellipse or of finite limited regularity.

10

10

f(x)=|x|, =-0.5, =-0.5

0

10

0

f(x)=|x|, =-0.9, =-0.8

||f(x)-H2n-1(f,x)||

||f(x)-H

10-0.2 log(n)n-1

10

-1

10

-0.2

2n-1

100

log(n)n

f(x)=|x|, =-0.3, =-0.6 ||f(x)-H

(f,x)||

n

-1

-1

10

2n-1 2max{ , }

10

2

(f,x)||

f(x)=|x|, =-0.2, =0.3 ||f(x)-H2n-1(f,x)|| n

2max{ , }

-1

101 10

-2

10

-3

10

1

10

2

n=11:2:1000

10

3

10

-2

10

-3

101

10-2

102

n=11:2:1000

103

10

-3

101

102

n=11:2:1000

103

10

0

101

102

103

n=11:2:1000

Fig. 2 H2n−1 (f, x) − f (x)∞ at x = −1 : 0.001 : 1 by using Gauss-Jacobi pointsystem for f (x) = |x| with different α and β, respectively

On the Convergence Rate of Hermite-Fejér Interpolation

627 (n)

(n)

(n)

For simplicity, in the following we abbreviate xk as xk , k (x) as k (x), hk (x) (n) as hk (x), and bk (x) as bk (x). A ∼ B denotes there exist two positive constants c1 and c2 such that c1 ≤ |A|/|B| ≤ c2 .

2 Main Results Suppose f (x) satisfies a Dini-Lipschitz condition on [−1, 1], then it has the following absolutely and uniformly convergent Chebyshev series expansion f (x) =

∞ 



cj Tj (x),

cj =

j =0

2 π



1 −1

f (x)Tj (x) √ dx, 1 − x2

j = 0, 1, . . . .

(12)

where the prime denotes summation whose first term is halved, Tj (x) = cos(j cos−1 x) denotes the Chebyshev polynomial of degree j . Lemma 1 (i) (Bernstein [2]) If f is analytic with |f (z)| ≤ M in the region bounded by the ellipse Eρ with foci ±1 and major and minor semiaxis lengths summing to ρ > 1, then for each j ≥ 0, |cj | ≤

2M . ρj

(13)

(ii) (Trefethen [13]) For an integer r ≥ 1, if f (x) has an absolutely continuous (r − 1)st derivative f (r−1) on [−1, 1] and a rth derivative f (r) of bounded variation Vr = Var(f (r) ) < ∞, then for each j ≥ r + 1, |cj | ≤

2Vr . πj (j − 1) · · · (j − r)

(14)

Suppose −1 < xn < xn−1 < · · · < x1 < 1 in decreasing order are the roots of (α,β) (x) (α, β > −1), and {wj }nj=1 are the corresponding weights in the GaussPn Jacobi quadrature. Lemma 2 For j = 1, 2, . . . , n, it follows * (x − xj )j (x) = σn (−1)j

(1 − xj2 )wj

2(α+β+1)/2



n!(n + α + β + 1) P (α,β) (x), (n + α + 1)(n + β + 1) n (15)

where σn = +1 for even n and σn = −1 for odd n.

628

S. Xiang and G. He

Proof Let zn = of

(α,β) Pn (x).

zn =

;1

−1 (1 − x)

α (1 + x)β [P (α,β) (x)]2 dx n

and Kn the leading coefficient

From Abramowitz and Stegun [1], we have

2α+β+1 (n + α + 1)(n + β + 1) · , 2n + α + β + 1 n!(n + α + β + 1)

Kn =

1 (2n + α + β + 1) . 2n n!(n + α + β + 1)

Furthermore, by Szegö [12, (15.3.1)] (also see Wang et al. [15]), we obtain  Kn2 2n(1 − xj2 )wj 1 ωn (x) = σn (−1)j (x − xj )j (x) =  ωn (x) ωn (xj )  2n(2n + α + β + 1)zn (1 − xj2 )wj P (α,β) (x), = σn (−1)j zn (2n + α + β + 1) n & %

which implies the desired result (15). Lemma 3 For j = 1, 2, . . . , n, it follows   (1 − xj2 )wj = O n−1 . 

(16)

2α+1   θj 2β+1 Proof From wj = O sin cos 2 Szegö [12, (15.3.10)],

     θj 2α+3 θj 2β+3 2α+β+3 π 2 , sin 2 cos 2 we see for xj = cos θj that (1−xj )wj = O n 2α+β+1 π n



θj 2

& %

which derives the desired result.

Lemma 4 ([10, 16]) For t ∈ [−1, 1], let xm be the root of the Jacobi polynomial (α,β) Pn which is closest to t. Then for k = 1, 2, . . . , n, we have ⎧   ⎨ O |k − m|−1 + |k − m|γ − 12 , k =  m , k (t) = ⎩ O(1) k=m

γ = max{α, β}.

(17)

Lemma 5 (Szegö [12, Theorem 8.1.2]) Let α, β be real but not necessarily greater than −1 and xk = cos θk . Then for each fixed k, it follows lim nθk = jk ,

n→∞

(18)

where jk is the kth positive zero of Bessel function Jα . Lemma 6 For k = 1, 2, . . . , n, it follows vk (x) = 1 − (x − xk )

ωn (xk ) = O(n2 ). ωn (xk )

(19)

On the Convergence Rate of Hermite-Fejér Interpolation

629

(α,β)

Proof Note that Pn (x) satisfies the second order linear homogeneous SturmLiouville differential equation [12, (4.2.1)] (1 − x 2 )y  + (β − α − (α + β + 2)x)y  + n(n + α + β + 1)y = 0. By ωn (x) =

(α,β)

Pn

(x)

Kn

, we get

β − α − (α + β + 2)xj ωn (xj ) =− ([12, (14.5.1)]). ωn (xj ) 1 − xj2 In addition, by Lemma 5 with xj = cos θj , we see that θ1 ∼ (α,β) Pn (−x)

=

(β,α) (−1)n Pn (x)

1 = O(n2 ), 1 − x12

we have θn ∼

1 = O(n2 ), 1 − xn2

1 n.

1 n.

(20)

Similarly, by

These together yield

# $ 1 1 1 ≤ max , = O(n2 ) 1 − xj2 1 − x12 1 − xn2

& %

and then by (20) it deduces the desired result. (α,β)

Theorem 1 Suppose {xj }nj=1 are the roots of Pn (x) with α, β > −1, then the Hermite-Fejér interpolation (5) for f analytic in Eρ with |f (z)| ≤ M at {xj }nj=1 has the convergence rate (9). Proof Since the Chebyshev series expansion of f (x) is uniformly convergent under the assumptions, and the error of Hermite-Fejér interpolation (5) on Chebyshev ∗ (Tj , x)| = 0 for j = polynomials satisfies |E(Tj , x)| = |Tj (x) − H2n−1 0, 1, . . . , 2n − 1, then it yields ∗ |E(f, x)| = |f (x) − H2n−1 (f, x)| = |

∞ 

∞ 

cj E(Tj , x)| ≤

j =0

|cj ||E(Tj , x)|.

j =2n

0: From |E(Tj , x)| = |Tj (x) − ni=1 Tj (xi )hi (x) − − 12 ,

(30)

while by Lemmas 2–3, we get ⎧   ⎪ ⎨ O log n , γ ≤ − 1 2  n | (xj − t)0+ (x − xj )2j (x)| ≤ |(x − xj )2j (x)| = 2γ , γ > − 1 . ⎪ O n ⎩ j =1 j =1 2 n 

n 

(31) Together (30) and (31), we can obtain the desired results by using ⎧   ⎪ ⎨ O log n , γ ≤ − 1 n 2   K2 (t) = . 2γ ⎪ , γ > − 12 . ⎩O n 1 Finally, We use a function of analytic f (x) = 1+25x 2 and a function of limited 5 ∗ regularity f (x) = |x| to show that the convergence rate of f (x)−H2n−1 (f, x)∞ is dependent on α and β in Fig. 3.

634

10

S. Xiang and G. He

5

f(x)=1/(1+25x2) =-0.7, =-0.3, = 0.3, = 0.6,

100

10

10

10

=-0.8 =-0.4 = 0.4 = 0.5

10

-10

10

10

1

10

n=1:10:1000

2

10

=-0.9, =-0.2, = 0.3, = 1.1,

100

-5

10-15 0 10

f(x)=|x|5

5

3

=-0.8 =-0.1 = 0.6 = 0.9

-5

-10

10-15 0 10

10

1

10

2

10

3

n=1:10:1000

∗ Fig. 3 H2n−1 (f, x) − f (x)∞ at x = −1 : 0.001 : 1 by using Gauss-Jacobi pointsystem for 1 5 f (x) = 1+25x 2 and f (x) = |x| with different α and β, respectively

Acknowledgements This author “Shuhuang Xiang” was supported partly by NSF of China (No. 11771454). This author “Guo He” was supported partly by NSF of China (No. 11901242), the Fundamental Research Funds for the Central Universities (No. 21618333), and the Opening Project at the Sun Yat-sen University (No. 2018010).

References 1. Abramowitz, M., Stegun, I.A.: Handbook of Mathematical Functions. National Bureau of Standards, Washington (1964) 2. Bernstein, S.: Sur l’ordre de la meilleure approximation des fonctions continues par les polynômes de degré donné. Mem. Cl. Sci. Acad. Roy. Belg. 4, 1–103 (1912) 3. Faber, G.: Über die interpolatorische Darstellung stetiger Funktionen. Jahresber. Deut. Math. Verein. 23, 192–210 (1914) 4. Fejér, L.: Über Interpolation, Nachrichten der Gesellschaft der Wissenschaften zu Göttingen Mathematisch-physikalische Klasse, 66–91 (1916) 5. Fejér, L.: Lagrangesche interpolation und die zugehörigen konjugierten Punkte. Math. Ann. 106, 1–55 (1932) 6. Fejér, L.: Bestimmung derjenigen Abszissen eines Intervalles, für welche die Quadratsumme der Grundfunktionen der Lagrangeschen Interpolation im Intervalle ein Möglichst kleines Maximum Besitzt. Ann. della Sc. Norm. Super. di Pisa 1, 263–276 (1932) 7. Grünwald, G.: On the theory of interpolation. Acta Math. 75, 219–245 (1942) 8. Kowalewski, G.: Interpolation und Genäherte Quadratur. Teubner-Verlag, Leipzig (1932) 9. Peano, G.: Resto nelle formule di quadrature, espresso con un integrale definito. Rom. Acc. L. Rend. 22, 562–569 (1913) 10. Sun, X.: Lagrange interpolation of functions of generalized bounded variation. Acta Math. Hungar. 53, 75–84 (1989) 11. Szabados, J.: On the order of magnitude of fundamental polynomials of Hermite interpolation. Acta Math. Hungar. 61, 357–368 (1993) 12. Szegö, G.: Orthogonal Polynomials, vol. 23. Colloquium Publications, American Mathematical Society, Providence (1939)

On the Convergence Rate of Hermite-Fejér Interpolation

635

13. Trefethen, L.N.: Approximation Theory and Approximation Practice. SIAM, Philadelphia (2012) 14. Vértesi, P.: Notes on the Hermite-Fejér interpolation based on the Jacobi abscissas. Acta Math. Acad. Sci. Hung. 24, 233–239 (1973) 15. Wang, H., Huybrechs, D., Vandewalle, S.: Explicit barycentric weights for polynomial interpolation in the roots or extrema of classical orthogonal polynomials. Math. Comput. 290, 2893–2914 (2012) 16. Xiang, S: On interpolation approximation: convergence rates for interpolation for functions of limited regularity. SIAM J. Numer. Anal. 54, 2081–2113 (2016) 17. Xiang, S., He, G.: The fast implementation of higher order Hermite-Fejér interpolation. SIAM J. Sci. Comput. 37, A1727–A1751 (2015) 18. Xiang, S., Chen, X., Wang, H: Error bounds for approximation in Chebyshev points. Numer. Math. 116, 463–491 (2010)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Fifth-Order Finite-Volume WENO on Cylindrical Grids Mohammad Afzal Shadab, Xing Ji, and Kun Xu

1 Introduction The conventional WENO scheme is specifically designed for the reconstruction in Cartesian coordinates on uniform grids [1]. The employment of Cartesian-based reconstruction scheme on a cylindrical grid suffers from a number of drawbacks [2, 3], e.g., in the original PPM paper, reconstruction was performed in volume coordinates (than the linear ones) so that algorithm for a Cartesian mesh can be used on a curvilinear mesh. However, the resulting interface states became first-order accurate even for smooth flows [2]. Another example can be the volume average assignment to the geometrical cell center of finite-volume than the centroid [2]. A breakthrough in the field of high order reconstruction in cylindrical coordinates is the application of the Vandermonde-like linear systems of equations with spatially

M. A. Shadab () Department of Mechanical and Aerospace Engineering, Hong Kong University of Science and Technology, Kowloon City, Hong Kong X. Ji Department of Mathematics, Hong Kong University of Science and Technology, Kowloon City, Hong Kong e-mail: [email protected] K. Xu Department of Mathematics, Hong Kong University of Science and Technology, Kowloon City, Hong Kong Department of Mechanical and Aerospace Engineering, Hong Kong University of Science and Technology, Kowloon City, Hong Kong e-mail: [email protected] © The Author(s) 2020 S. J. Sherwin et al. (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2018, Lecture Notes in Computational Science and Engineering 134, https://doi.org/10.1007/978-3-030-39647-3_51

637

638

M. A. Shadab et al.

varying coefficients [2]. It is reintroduced in the present work to build a basis for the derivation of the high order WENO schemes. The motivation for the present work is to develop a fifth-order finite-volume WENO reconstruction scheme in the efficient dimension-by-dimension framework, specifically aimed at regularly-spaced and irregularly-spaced grids in cylindrical coordinates.

2 Finite-Volume Discretization in Curvilinear Coordinates 2.1 Evaluation of the Linear Weights A non-uniform grid spacing with zone width Δξi = ξi+ 1 − ξi− 1 is considered 2 2 having ξ ∈ (x1 , x2 , x3 ) as the coordinate along the reconstruction direction and ξi+ 1 denoting the location of the cell interface between zones i and i + 1. Let Q¯ i 2 be the cell average of conserved quantity Q inside zone i at some given time, which can be expressed in form of Eq. (1). Q¯ i =

1 ΔVi



ξi+ 1

2

ξi− 1

Qi (ξ )

∂V dξ ∂ξ

 &

ΔVi =

ξi+ 1

2

ξi− 1

2

∂V dξ ∂ξ

(1)

2

where the local cell volume ΔVi of ith cell in the direction of reconstruction given in Eq. (1) and ∂∂ξV is the one-dimensional Jacobian. Now, our aim is to find a pth order accurate approximation to the actual solution by constructing a (p − 1)th order polynomial distribution, as given in Eq. (2). Qi (ξ ) = ai,0 + ai,1 (ξ − ξic ) + ai,2 (ξ − ξic )2 + . . . + ai,p−1 (ξ − ξic )p−1

(2)

where ai,n corresponds to a vector of the coefficients which needs to be determined and ξic can be taken as the cell centroid. However, the final values at the interface are independent of the particular choice of ξic and one may as well set ξic = 0 [2]. Unlike the cell center, the centroid is not equidistant from the cell interfaces in the case of cylindrical-radial coordinates, and the cell averaged values are assigned at the centroid [2]. Further, the method has to be locally conservative, i.e., the polynomial Qi (ξ ) must fit the neighboring cell averages, satisfying Eq. (3). 

ξi+s+ 1

2

ξi+s− 1

Qi (ξ )

∂V dξ = ΔVi+s Q¯ i+s ∂ξ

for

− iL ≤ s ≤ iR

(3)

2

where the stencil includes iL cells to the left and iR cells to the right of the ith zone such that iL + iR + 1 = p. Implementing Eqs. (1)–(2) in Eq. (3) along with a simple mathematical manipulation leads to Eq. (4), which is the fundamental equation for

Fifth-Order Finite-Volume WENO on Cylindrical Grids

639

reconstruction in cylindrical coordinates. For the detailed derivation, kindly refer to [3]. ⎛

⎞T βi−iL ,0 . . . βi−iL ,p−1 ⎜ . ⎟ .. .. ⎜ . ⎟ . . . ⎝ ⎠ βi+iR ,0 . . . βi+iR ,p−1

⎛ ⎞ ⎛ ± wi,−i ⎜ . L⎟ ⎜ ⎜ . ⎟=⎜ ⎝ . ⎠ ⎝ ± (ξi± 1 wi,i R 2

1 .. . − ξic )p−1

⎞ ⎟ ⎟ ⎠

(4)

where ‘±’ represents the positive and negative weights i.e. weights for reconstructing right (+) and left (−) interface values respectively. Also, the grid ± dependent linear weights (wi,s ) satisfy the normalization condition [2].

2.2 Optimal Weights For the case of fifth-order WENO interpolation, the third order interpolated variables are optimally weighed in order to achieve fifth-order accurate interpolated values as given in Eq. (5) for the case of p = 3 [1]. (2p−1)±

qi,0

=

p−1 

± Ci,l qi,l

p±

(5)

l=0 ± where Ci,l is the optimal weight for the positive/negative cases on the ith finitevolume. So, Eq. (4) is used again to evaluate the weights for the fifth-order (2p−1 = 5) interpolation (iL = 2, iR = 2). Linear and optimal weights are independent of the mesh size for standard regularly-spaced grid cases. They can be evaluated and stored (at a nominal cost) independently before the actual computation. Also, they conform to the original WENO-JS [1] for the limiting case (R → ∞). The weights required for source term and flux integration in one or more dimensions are given in [3].

2.3 Smoothness Indicators and the Nonlinear Weights The mathematical definition of the smoothness indicator is given in Eq. (6) [1]. I Si,l =

p−1   ξj+ 1

2

m=1 ξj− 12



dm Qi,l (ξ ) dξ m

2 Δξi2m−1 dξ ,

l = 0, . . . , p − 1

(6)

To evaluate the value of I Si,l , a third order polynomial interpolation on ith cell is required using positive and negative reconstructed values by stencil Sl , as given in Eq. (2). Finally, evaluating the values of the coefficient a’s and substituting their

640

M. A. Shadab et al.

values in smoothness indicator formula (6) yields the grid-independent fundamental ± relation (7). The nonlinear weight (ωi,l ) for the WENO-C interpolation is defined in Eq. (8) [1], where # is chosen to be 10−6 [1, 3]. ¯ 2i − 39Q ¯ i (q − + q + ) + 10((q − )2 + (q + )2 ) + 19q − q + ) I Si,l = 4(39Q i,l i,l i,l i,l i,l i,l ± αi,l ± ωi,l = 0 (10) ∂t (∂V/∂ξ ) ∂ξ ∂ξ where Q is the conserved variable, (∂V/∂ξ ) = ξ is the one-dimensional Jacobian in cylindrical-radial coordinates. Boundary conditions are not considered in the present approach to reduce the complexity of the analysis. Assuming a uniform grid with ξi = iΔξ and ξi+1 − ξi = Δξ ∀i and (i ± 1/2) denotes the boundaries of the finitevolume i. In the finite-volume framework, Eq. (10) transforms into the conservative scheme given in Eq. (11). ¯i ∂Q 1 ˆ =− (Fi+1/2 − Fˆi−1/2 ) ∂t ΔVi

(11)

where numerical flux Fˆi+1/2 is the Lax-Friedrich flux, and Q¯ i and Vi are given in Eq. (1). For this particular problem, let v = 1 in Eq. (10). Therefore, only the values on the left side of the interface are considered. Based on the von Neumann stability analysis, the semi-discrete solution can be expressed as a discrete Fourier series. By the superposition principle, only one term in the series can be used for analysis, as

Fifth-Order Finite-Volume WENO on Cylindrical Grids

641

illustrated in Eq. (12). Q¯ i (t) = Qˆ k (t)ej iθk ,

where j =

√ −1

(12)

By substituting Eq. (12) in Eq. (11), we can separate the spatial operator L, as given in Eq. (13). L=−

− [Q(∂ V/∂ξ )]− (Fˆi+1/2 − Fˆi−1/2 ) z(θk )Q¯ i i+1/2 − [Q(∂ V/∂ξ )]i−1/2 =− =− ΔVi ΔVi Δξ (13)

where the complex function z(θk ) is the Fourier symbol. By substituting the values − of Q− i−1/2 and Qi+1/2 using fifth-order positive weights of cells (i − 1) and i respectively for a smooth solution, the value of z(θk ) for WENO-C can be evaluated using Eq. (14). +2   m+1 + m j lθk + z(θk ) = (m+1) i e − w(i−1),l (i − 1)m ej (l−1)θk wi,l i − (i − 1)(m+1) l=−2

(14) where m = 1 for cylindrical-radial coordinates. Using the same approach as given in [4], we can plot the spatial spectrum {S : −z(θk ) for θk ∈ [0, 2π]} and the stability domain St for TVD-RK order 3. The maximum stable CFL number of this scheme can be computed by finding the largest rescaling parameter σ˜ , so that the rescaled spectrum still lies in the stability domain. It can be observed from Fig. 1 that the spatial spectrums S of WENO-C differs initially with the index numbers i due to the geometrical variation of the finitevolume. However, the spectrums are the same for high index numbers (i), similar to WENO-JS, as the fifth-order interpolation weights converge. Some regions (i = 1, 2) require boundary conditions and thus, are not considered in the present analysis. The values of CFL number for cylindrical-radial coordinates lie in between 1.45 and 1.52. As a final remark, it can be concluded that the proposed scheme is A-stable with third or higher order of RK method with an appropriate value of CFL number for this case.

4 Numerical Tests In this section, several tests on Euler equations are performed to analyze the performance of the WENO-C reconstruction scheme. Tests are performed on a gamma law gas (γ = 1.4) in cylindrical coordinates to investigate the essentially non-oscillatory property of WENO-C for discontinuous flows and the convex combination property for smooth flows. For first-order and second-order (MUSCL)

Fig. 1 Rescaled spectrums (with maximum stable CFL number σ˜ ) and stability domains of fifth-order WENO-C in cylindrical-radial coordinates in a complex plane for different cell index numbers i. (a) i = 3, σ˜ = 1.45. (b) i = 5, σ˜ = 1.52. (c) i = 10, σ˜ = 1.50. (d) i = 50, σ˜ = 1.46. (e) i = 100, σ˜ = 1.45. (f) Legend

642 M. A. Shadab et al.

Fifth-Order Finite-Volume WENO on Cylindrical Grids

643

spatial reconstructions, Euler time marching and Maccormack (predictor-corrector) schemes are respectively employed. For WENO-C, time marching is done with TVD-RK order 3 for 1D cases and RK order 5 for the 2D case.

4.1 Acoustic Wave Propagation A smooth problem involving a nonlinear system of 1D gas dynamical equations is solved to test fifth-order accuracy of the spatial discretization scheme [3]. The Euler equations in cylindrical-radial coordinates can be written in the form of Eq. (15). ⎛

⎛ ⎞ ⎛ ⎞ ⎞ ρ ρuR 0 ∂ ⎜ ⎟ 1 ∂ ⎜ ⎟ ⎜ ⎟ ⎝ρu⎠ + ⎝(ρu2 + p)R ⎠ = ⎝p/R ⎠ ∂t R ∂R (E + p)uR E 0

(15)

where ρ is the mass density, u is the radial velocity, p is the pressure, and E is the total energy. Equation (16) serves as the adiabatic equation of state. E=

1 p + ρu2 γ −1 2

(16)

The initial conditions are provided in Eq. (17) with the perturbation given in Eq. (18). The interface flux is evaluated with Rusanov scheme [3]. ρ(R, 0) = 1 + εf (R), f (R) =

u(R, 0) = 0,

p(R, 0) = 1/γ + εf (R)

⎧ ⎨ sin4 (5πR)

if 0.4 ≤ R ≤ 0.6

⎩0

otherwise

R

(17)

(18)

A sufficiently small perturbation with ε = 10−4 yields a smooth solution. The interface flux is evaluated using Rusanov scheme with a CFL number of 0.3. The initial perturbation splits into two acoustic waves traveling in opposite directions. The final time (t = 0.3) is set such that the waves remain in the domain and the problem is free from the boundary effects. The computational domain of unity length is uniformly divided into N different zones i.e. N = 16, 32, 64, 128, 256. Although an exact solution known up to O(ε2 ) is known, the solution on the finest mesh N = 1024 is taken as the reference. Figure 2 illustrate the spatial variation of density at t = 0.3 inside the domain. From Table 1, it clear that the scheme approaches the desired fifth-order accuracy.

644

M. A. Shadab et al.

1.00016

N = 16 N = 32 N = 64 N = 128 N = 256 N = 1024

1.00014 1.00012

Density

1.0001 1.00008 1.00006 1.00004 1.00002 1 0

0.2

0.4

R

0.6

0.8

1

Fig. 2 Spatial profiles of density at t = 0.3 for acoustic wave propagation test in cylindrical-radial coordinates Table 1 L1 norm errors and order of convergence table for acoustic wave propagation test

N 16 32 64 128 256

#(ρ) 1.01E−05 4.91E−06 6.74E−07 3.24E−08 1.27E−09

OL1 – 1.036 2.865 4.380 4.670

4.2 Sedov Explosion Test Sedov explosion test is performed to investigate code’s ability to deal with strong shocks and non-planar symmetry [3]. The problem involves a self-similar evolution of a cylindrical blastwave in a uniform grid (N = 100) from a localized initial pressure perturbation (delta-function) in an otherwise homogeneous medium. Governing equations are given in Eq. (15) and the fluxes are evaluated with Rusanov scheme and GKS [5]. For the code initialization, dimensionless energy # = 1

Fifth-Order Finite-Volume WENO on Cylindrical Grids

645

is deposited into a small region of radius δ = 3ΔR. Inside this region, the  dimensionless pressure P0 is given by Eq. (19). 

P0 =

3(γ − 1)# (m + 2)πδ (m+1)

(19)

where m = 1 for cylindrical geometry. Reflecting boundary condition is employed at the center (R = 0), whereas boundary condition at R = 1 is not required for this problem. The initial velocity and density inside the domain are 0 and 1 respectively and the initial pressure everywhere except the kernel is 10−5 . As the source term is very stiff, the CFL number is set to be 0.1. The final time is t = 0.05. Figure 3 shows that the peak for WENO-C is higher for density and is closest to the analytical value, similar to fifth-order finite difference version [3], but MUSCL has higher offset peaks for pressure and velocity. GKS performs slightly better than RS, as the peaks are slightly higher for all the cases.

6

Analytical First Order MUSCL WENO-C (RS) WENO-C (GKS)

5

3 2

0

0. 1

0. 2 R

1

0. 3

0. 4

0

0

0. 1

0. 2 R

0. 3

0. 4

Analytical First Order MUSCL WENO-C (RS) WENO-C (GKS)

4 3 Pressure

1. 5

0. 5

1 0

Analytical First Order MUSCL WENO-C (RS) WENO-C (GKS)

2

Velocity

Density

4

2. 5

2 1 0

0

0. 1

0. 2 R

0. 3

0. 4

Fig. 3 Variation of density, velocity, and pressure with the radius for Sedov explosion test in cylindrical-radial coordinates. Domain is restricted to R = 0.4 for the sake of clarity

646

M. A. Shadab et al.

4.3 Modified 2D Riemann Problem in (R − z) Coordinates

Radial symmetry

The final test for the present scheme involves a modified 2D Riemann problem in cylindrical (R − z) coordinates, as illustrated in Fig. 4 (top left). The problem involves 2 contact discontinuities and 2 shocks as the initial condition, resulting in the formation of a self-similar structure propagating towards the low densitylow pressure region (region 3). The governing equations in cylindrical (R − z) coordinates are provided in Eq. (20). The computations are performed until t = 0.2 with a CFL number of 0.5 on a domain (R, z)=[0,1]×[0,1] divided into 500×500 zones. The boundary conditions are symmetry at the center (except for the antisymmetric radial velocity) and outflow elsewhere. HLL Riemann solver is used for flux evaluations. Rich smallscale structures in the contact-contact region (region 1) can be observed from

=1 =1 4 = 0.51 4 = −0.51

Z

4 4

=1 1 = 0.8 1 =0 1 =0

1

+ +

=1 2 =1 2 = 0.51 2 = 0.51

2

1

= 0.4 = 0.53 3 3 =0 3 = 3

0. 8 0. 6 0. 4 0. 2 0

0. 4

0. 6

0. 8

0. 2

0. 4

0. 6

0. 8

1

R

Z

Z

1

1

0. 8

0. 8

0. 6

0. 6

0. 4

0. 4

0. 2

0. 2 0

0. 2

0. 2

0. 4

0. 6

0. 8

1

R

0

1

R

Fig. 4 Modified 2D Riemann problem in cylindrical (r − z) coordinates: schematic (top left), density contours at t = 0.2 with first-order (top right), second-order MUSCL (bottom left), and WENO-C (bottom right) reconstruction schemes

Fifth-Order Finite-Volume WENO on Cylindrical Grids

647

Fig. 4 for WENO-C reconstruction, when compared with first and second-order MUSCL reconstruction. Structures are highly smeared for the case of first-order reconstruction. ⎞ ⎛ ⎛ ⎛ ⎛ ⎞ ⎞ ⎞ ρ ρvR R ρvz 0 ⎟ ⎜ ⎜ ⎟ ⎟ ⎟ ∂ ⎜ ∂ ⎜ ⎜ρvR ⎟ 1 ∂ ⎜ (ρvR2 + p)R ⎟ ⎜ ρvR vz ⎟ ⎜p/R ⎟ (20) ⎟=⎜ ⎜ ⎜ ⎜ ⎟+ ⎟+ ⎟ 2 ∂t ⎝ ρvz ⎠ R ∂R ⎝ ρvR vz R ⎠ ∂z ⎝ ρvz + p ⎠ ⎝ 0 ⎠ (ρe + p)vR R (ρe + p)vz ρe 0

5 Conclusions The fifth-order finite-volume WENO-C reconstruction scheme is proposed for structured grids in cylindrical coordinates to achieve high order spatial accuracy along with ENO transition. A grid independent smoothness indicator is derived for this scheme. For uniform grids, the analytical values in cylindrical-radial coordinates for the limiting case (R → ∞) conform to WENO-JS. Linear stability analysis of the present scheme is performed using a scalar advection equation in radial coordinates. Several tests involving smooth and discontinuous flows are performed, which testify for the fifth-order accuracy and ENO property of the scheme.

References 1. Jiang, G.-S., Shu, C.-W.: Efficient implementation of weighted ENO schemes. J. Comput. Phys. 126(1), 202–228 (1996) 2. Mignone, A.: High-order conservative reconstruction schemes for finite volume methods in cylindrical and spherical coordinates. J. Comput. Phys. 270, 784–814 (2014) 3. Shadab, M.A., Balsara, D., Shyy, W., Xu, K.: Fifth order finite volume WENO in general orthogonally-curvilinear coordinates. Comput. Fluids. 190, 398–424 (2019) 4. Liu, H., Jiao, X.: WLS-ENO: Weighted-least-squares based essentially non-oscillatory schemes for finite volume methods on unstructured meshes. J. Comput. Phys. 314, 749–773 (2016) 5. Xu, K.: A gas-kinetic BGK scheme for the Navier–Stokes equations and its connection with artificial dissipation and Godunov method. J. Comput. Phys. 171(1), 289–335 (2001)

648

M. A. Shadab et al.

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Editorial Policy 1. Volumes in the following three categories will be published in LNCSE: i) Research monographs ii) Tutorials iii) Conference proceedings Those considering a book which might be suitable for the series are strongly advised to contact the publisher or the series editors at an early stage. 2. Categories i) and ii). Tutorials are lecture notes typically arising via summer schools or similar events, which are used to teach graduate students. These categories will be emphasized by Lecture Notes in Computational Science and Engineering. Submissions by interdisciplinary teams of authors are encouraged. The goal is to report new developments – quickly, informally, and in a way that will make them accessible to non-specialists. In the evaluation of submissions timeliness of the work is an important criterion. Texts should be well-rounded, well-written and reasonably self-contained. In most cases the work will contain results of others as well as those of the author(s). In each case the author(s) should provide sufficient motivation, examples, and applications. In this respect, Ph.D. theses will usually be deemed unsuitable for the Lecture Notes series. Proposals for volumes in these categories should be submitted either to one of the series editors or to Springer-Verlag, Heidelberg, and will be refereed. A provisional judgement on the acceptability of a project can be based on partial information about the work: a detailed outline describing the contents of each chapter, the estimated length, a bibliography, and one or two sample chapters – or a first draft. A final decision whether to accept will rest on an evaluation of the completed work which should include – at least 100 pages of text; – a table of contents; – an informative introduction perhaps with some historical remarks which should be accessible to readers unfamiliar with the topic treated; – a subject index. 3. Category iii). Conference proceedings will be considered for publication provided that they are both of exceptional interest and devoted to a single topic. One (or more) expert participants will act as the scientific editor(s) of the volume. They select the papers which are suitable for inclusion and have them individually refereed as for a journal. Papers not closely related to the central topic are to be excluded. Organizers should contact the Editor for CSE at Springer at the planning stage, see Addresses below. In exceptional cases some other multi-author-volumes may be considered in this category. 4. Only works in English will be considered. For evaluation purposes, manuscripts may be submitted in print or electronic form, in the latter case, preferably as pdf- or zipped ps-files. Authors are requested to use the LaTeX style files available from Springer at http:// www.springer.com/gp/authors-editors/book-authors-editors/manuscript-preparation/5636 (Click on LaTeX Template → monographs or contributed books). For categories ii) and iii) we strongly recommend that all contributions in a volume be written in the same LaTeX version, preferably LaTeX2e. Electronic material can be included if appropriate. Please contact the publisher. Careful preparation of the manuscripts will help keep production time short besides ensuring satisfactory appearance of the finished book in print and online.

5. The following terms and conditions hold. Categories i), ii) and iii): Authors receive 50 free copies of their book. No royalty is paid. Volume editors receive a total of 50 free copies of their volume to be shared with authors, but no royalties. Authors and volume editors are entitled to a discount of 40 % on the price of Springer books purchased for their personal use, if ordering directly from Springer. 6. Springer secures the copyright for each volume. Addresses:

Timothy J. Barth NASA Ames Research Center NAS Division Moffett Field, CA 94035, USA [email protected] Michael Griebel Institut für Numerische Simulation der Universität Bonn Wegelerstr. 6 53115 Bonn, Germany [email protected] David E. Keyes Mathematical and Computer Sciences and Engineering King Abdullah University of Science and Technology P. O. Box 55455 Jeddah 21534, Saudi Arabia [email protected] and Department of Applied Physics and Applied Mathematics Columbia University 500 W. 120th Street New York, NY 10027, USA [email protected]

Risto M. Nieminen Department of Applied Physics Aalto University School of Science and Technology 00076 Aalto, Finland [email protected] Dirk Roose Department of Computer Science Katholieke Universiteit Leuven Celestijnenlaan 200A 3001 Leuven-Heverlee, Belgium [email protected] Tamar Schlick Department of Chemistry and Courant Institute of Mathematical Sciences New York University 251 Mercer Street New York, NY 10012, USA [email protected] Editor for Computational Science and Engineering at Springer: Martin Peters Springer-Verlag Mathematics Editorial IV Tiergartenstrasse 17 69121 Heidelberg, Germany [email protected]

Lecture Notes in Computational Science and Engineering 1. D. Funaro, Spectral Elements for Transport-Dominated Equations. 2. H.P. Langtangen, Computational Partial Differential Equations. Numerical Methods and Diffpack Programming. 3. W. Hackbusch, G. Wittum (eds.), Multigrid Methods V . 4. P. Deuflhard, J. Hermans, B. Leimkuhler, A.E. Mark, S. Reich, R.D. Skeel (eds.), Computational Molecular Dynamics: Challenges, Methods, Ideas. 5. D. Kröner, M. Ohlberger, C. Rohde (eds.), An Introduction to Recent Developments in Theory and Numerics for Conservation Laws. 6. S. Turek, Efficient Solvers for Incompressible Flow Problems. An Algorithmic and Computational Approach. 7. R. von Schwenn, Multi Body System SIMulation. Numerical Methods, Algorithms, and Software. 8. H.-J. Bungartz, F. Durst, C. Zenger (eds.), High Performance Scientific and Engineering Computing. 9. T.J. Barth, H. Deconinck (eds.), High-Order Methods for Computational Physics. 10. H.P. Langtangen, A.M. Bruaset, E. Quak (eds.), Advances in Software Tools for Scientific Computing. 11. B. Cockburn, G.E. Karniadakis, C.-W. Shu (eds.), Discontinuous Galerkin Methods. Theory, Computation and Applications. 12. U. van Rienen, Numerical Methods in Computational Electrodynamics. Linear Systems in Practical Applications. 13. B. Engquist, L. Johnsson, M. Hammill, F. Short (eds.), Simulation and Visualization on the Grid. 14. E. Dick, K. Riemslagh, J. Vierendeels (eds.), Multigrid Methods Vl. 15. A. Frommer, T. Lippert, B. Medeke, K. Schilling (eds.), Numerical Challenges in Lattice Quantum Chromodynamics. 16. J. Lang, Adaptive Multilevel Solution of Nonlinear Parabolic PDE Systems. Theory, Algorithm, and Applications. 17. B.I. Wohlmuth, Discretization Methods and Iterative Solvers Based on Domain Decomposition. 18. U. van Rienen, M. Günther, D. Hecht (eds.), Scientific Computing in Electrical Engineering. 19. I. Babuška, P.G. Ciarlet, T. Miyoshi (eds.), Mathematical Modeling and Numerical Simulation in Continuum Mechanics. 20. T.J. Barth, T. Chan, R. Haimes (eds.), Multiscale and Multiresolution Methods. Theory and Applications. 21. M. Breuer, F. Durst, C. Zenger (eds.), High Performance Scientific and Engineering Computing. 22. K. Urban, Wavelets in Numerical Simulation. Problem Adapted Construction and Applications. 23. L.F. Pavarino, A. Toselli (eds.), Recent Developments in Domain Decomposition Methods.

24. T. Schlick, H.H. Gan (eds.), Computational Methods for Macromolecules: Challenges and Applications. 25. T.J. Barth, H. Deconinck (eds.), Error Estimation and Adaptive Discretization Methods in Computational Fluid Dynamics. 26. M. Griebel, M.A. Schweitzer (eds.), Meshfree Methods for Partial Differential Equations. 27. S. Müller, Adaptive Multiscale Schemes for Conservation Laws. 28. C. Carstensen, S. Funken, W. Hackbusch, R.H.W. Hoppe, P. Monk (eds.), Computational Electromagnetics. 29. M.A. Schweitzer, A Parallel Multilevel Partition of Unity Method for Elliptic Partial Differential Equations. 30. T. Biegler, O. Ghattas, M. Heinkenschloss, B. van Bloemen Waanders (eds.), Large-Scale PDEConstrained Optimization. 31. M. Ainsworth, P. Davies, D. Duncan, P. Martin, B. Rynne (eds.), Topics in Computational Wave Propagation. Direct and Inverse Problems. 32. H. Emmerich, B. Nestler, M. Schreckenberg (eds.), Interface and Transport Dynamics. Computational Modelling. 33. H.P. Langtangen, A. Tveito (eds.), Advanced Topics in Computational Partial Differential Equations. Numerical Methods and Diffpack Programming. 34. V John, Large Eddy Simulation of Turbulent Incompressible Flows. Analytical and Numerical Results for a Class of LES Models. 35. E. Bänsch (ed.), Challenges in Scientific Computing- CISC 2002. 36. B.N. Khoromskij, G. Wittum, Numerical Solution of Elliptic Differential Equations by Reduction to the Interface. 37. A. Iske, Multiresolution Methods in Scattered Data Modelling. 38. S.-I. Niculescu, K. Gu (eds.), Advances in Time-Delay Systems. 39. S. Attinger, P. Koumoutsakos (eds.), Multiscale Modelling and Simulation. 40. R. Kornhuber, R. Hoppe, J. Périaux, O. Pironneau, O. Wildlund, J. Xu (eds.), Domain Decomposition Methods in Science and Engineering. 41. T. Plewa, T. Linde, V.G. Weirs (eds.), Adaptive Mesh Refinement – Theory and Applications. 42. A. Schmidt, K.G. Siebert, Design of Adaptive Finite Element Software. The Finite Element Toolbox ALBERTA. 43. M. Griebel, M.A. Schweitzer (eds.), Meshfree Methods for Partial Differential Equations II. 44. B. Engquist, P. Lötstedt, O. Runborg (eds.), Multiscale Methods in Science and Engineering. 45. P. Benner, V. Mehrmann, D.C. Sorensen (eds.), Dimension Reduction of Large-Scale Systems. 46. D. Kressner, Numerical Methods for General and Structured Eigenvalue Problems. 47. A. Boriçi, A. Frommer, B. Joó, A. Kennedy, B. Pendleton (eds.), QCD and Numerical Analysis III. 48. F. Graziani (ed.), Computational Methods in Transport. 49. B. Leimkuhler, C. Chipot, R. Elber, A. Laaksonen, A. Mark, T. Schlick, C. Schütte, R. Skeel (eds.), New Algorithms for Macromolecular Simulation.

50. M. Bücker, G. Corliss, P. Hovland, U. Naumann, B. Norris (eds.), Automatic Differentiation: Applications, Theory, and Implementations. 51. A.M. Bruaset, A. Tveito (eds.), Numerical Solution of Partial Differential Equations on Parallel Computers. 52. K.H. Hoffmann, A. Meyer (eds.), Parallel Algorithms and Cluster Computing. 53. H.-J. Bungartz, M. Schäfer (eds.), Fluid-Structure Interaction. 54. J. Behrens, Adaptive Atmospheric Modeling. 55. O. Widlund, D. Keyes (eds.), Domain Decomposition Methods in Science and Engineering XVI. 56. S. Kassinos, C. Langer, G. Iaccarino, P. Moin (eds.), Complex Effects in Large Eddy Simulations. 57. M. Griebel, M.A Schweitzer (eds.), Meshfree Methods for Partial Differential Equations III. 58. A.N. Gorban, B. Kégl, D.C. Wunsch, A. Zinovyev (eds.), Principal Manifolds for Data Visualization and Dimension Reduction. 59. H. Ammari (ed.), Modeling and Computations in Electromagnetics: A Volume Dedicated to JeanClaude Nédélec. 60. U. Langer, M. Discacciati, D. Keyes, O. Widlund, W. Zulehner (eds.), Domain Decomposition Methods in Science and Engineering XVII. 61. T. Mathew, Domain Decomposition Methods for the Numerical Solution of Partial Differential Equations. 62. F. Graziani (ed.), Computational Methods in Transport: Verification and Validation. 63. M. Bebendorf, Hierarchical Matrices. A Means to Efficiently Solve Elliptic Boundary Value Problems. 64. C.H. Bischof, H.M. Bücker, P. Hovland, U. Naumann, J. Utke (eds.), Advances in Automatic Differentiation. 65. M. Griebel, M.A. Schweitzer (eds.), Meshfree Methods for Partial Differential Equations IV. 66. B. Engquist, P. Lötstedt, O. Runborg (eds.), Multiscale Modeling and Simulation in Science. 67. I.H. Tuncer, Ü. Gülcat, D.R. Emerson, K. Matsuno (eds.), Parallel Computational Fluid Dynamics 2007. 68. S. Yip, T. Diaz de la Rubia (eds.), Scientific Modeling and Simulations. 69. A. Hegarty, N. Kopteva, E. O’Riordan, M. Stynes (eds.), BAIL 2008 – Boundary and Interior Layers. 70. M. Bercovier, M.J. Gander, R. Kornhuber, O. Widlund (eds.), Domain Decomposition Methods in Science and Engineering XVIII. 71. B. Koren, C. Vuik (eds.), Advanced Computational Methods in Science and Engineering. 72. M. Peters (ed.), Computational Fluid Dynamics for Sport Simulation. 73. H.-J. Bungartz, M. Mehl, M. Schäfer (eds.), Fluid Structure Interaction II - Modelling, Simulation, Optimization. 74. D. Tromeur-Dervout, G. Brenner, D.R. Emerson, J. Erhel (eds.), Parallel Computational Fluid Dynamics 2008. 75. A.N. Gorban, D. Roose (eds.), Coping with Complexity: Model Reduction and Data Analysis.

76. J.S. Hesthaven, E.M. Rønquist (eds.), Spectral and High Order Methods for Partial Differential Equations. 77. M. Holtz, Sparse Grid Quadrature in High Dimensions with Applications in Finance and Insurance. 78. Y. Huang, R. Kornhuber, O.Widlund, J. Xu (eds.), Domain Decomposition Methods in Science and Engineering XIX. 79. M. Griebel, M.A. Schweitzer (eds.), Meshfree Methods for Partial Differential Equations V. 80. P.H. Lauritzen, C. Jablonowski, M.A. Taylor, R.D. Nair (eds.), Numerical Techniques for Global Atmospheric Models. 81. C. Clavero, J.L. Gracia, F.J. Lisbona (eds.), BAIL 2010 – Boundary and Interior Layers, Computational and Asymptotic Methods. 82. B. Engquist, O. Runborg, Y.R. Tsai (eds.), Numerical Analysis and Multiscale Computations. 83. I.G. Graham, T.Y. Hou, O. Lakkis, R. Scheichl (eds.), Numerical Analysis of Multiscale Problems. 84. A. Logg, K.-A. Mardal, G. Wells (eds.), Automated Solution of Differential Equations by the Finite Element Method. 85. J. Blowey, M. Jensen (eds.), Frontiers in Numerical Analysis - Durham 2010. 86. O. Kolditz, U.-J. Gorke, H. Shao, W. Wang (eds.), Thermo-Hydro-Mechanical-Chemical Processes in Fractured Porous Media- Benchmarks and Examples. 87. S. Forth, P. Hovland, E. Phipps, J. Utke, A. Walther (eds.), Recent Advances in Algorithmic Differentiation. 88. J. Garcke, M. Griebel (eds.), Sparse Grids and Applications. 89. M. Griebel, M.A. Schweitzer (eds.), Meshfree Methods for Partial Differential Equations VI. 90. C. Pechstein, Finite and Boundary Element Tearing and Interconnecting Solvers for Multiscale Problems. 91. R. Bank, M. Holst, O. Widlund, J. Xu (eds.), Domain Decomposition Methods in Science and Engineering XX. 92. H. Bijl, D. Lucor, S. Mishra, C. Schwab (eds.), Uncertainty Quantification in Computational Fluid Dynamics. 93. M. Bader, H.-J. Bungartz, T. Weinzierl (eds.), Advanced Computing. 94. M. Ehrhardt, T. Koprucki (eds.), Advanced Mathematical Models and Numerical Techniques for Multi-Band Effective Mass Approximations. 95. M. Azaïez, H. El Fekih, J.S. Hesthaven (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM 2012. 96. F. Graziani, M.P. Desjarlais, R. Redmer, S.B. Trickey (eds.), Frontiers and Challenges in Warm Dense Matter. 97. J. Garcke, D. Pflüger (eds.), Sparse Grids and Applications – Munich 2012. 98. J. Erhel, M. Gander, L. Halpern, G. Pichot, T. Sassi, O. Widlund (eds.), Domain Decomposition Methods in Science and Engineering XXI. 99. R. Abgrall, H. Beaugendre, P.M. Congedo, C. Dobrzynski, V. Perrier, M. Ricchiuto (eds.), High Order Nonlinear Numerical Methods for Evolutionary PDEs - HONOM 2013. 100. M. Griebel, M.A. Schweitzer (eds.), Meshfree Methods for Partial Differential Equations VII.

101. R. Hoppe (ed.), Optimization with PDE Constraints - OPTPDE 2014. 102. S. Dahlke, W. Dahmen, M. Griebel, W. Hackbusch, K. Ritter, R. Schneider, C. Schwab, H. Yserentant (eds.), Extraction of Quantifiable Information from Complex Systems. 103. A. Abdulle, S. Deparis, D. Kressner, F. Nobile, M. Picasso (eds.), Numerical Mathematics and Advanced Applications - ENUMATH 2013. 104. T. Dickopf, M.J. Gander, L. Halpern, R. Krause, L.F. Pavarino (eds.), Domain Decomposition Methods in Science and Engineering XXII. 105. M. Mehl, M. Bischoff, M. Schäfer (eds.), Recent Trends in Computational Engineering - CE2014. Optimization, Uncertainty, Parallel Algorithms, Coupled and Complex Problems. 106. R.M. Kirby, M. Berzins, J.S. Hesthaven (eds.), Spectral and High Order Methods for Partial Differential Equations - ICOSAHOM’14. 107. B. Jüttler, B. Simeon (eds.), Isogeometric Analysis and Applications 2014. 108. P. Knobloch (ed.), Boundary and Interior Layers, Computational and Asymptotic Methods – BAIL 2014. 109. J. Garcke, D. Pflüger (eds.), Sparse Grids and Applications – Stuttgart 2014. 110. H. P. Langtangen, Finite Difference Computing with Exponential Decay Models. 111. A. Tveito, G.T. Lines, Computing Characterizations of Drugs for Ion Channels and Receptors Using Markov Models. 112. B. Karazösen, M. Manguo˘glu, M. Tezer-Sezgin, S. Göktepe, Ö. U˘gur (eds.), Numerical Mathematics and Advanced Applications - ENUMATH 2015. 113. H.-J. Bungartz, P. Neumann, W.E. Nagel (eds.), Software for Exascale Computing - SPPEXA 2013– 2015. 114. G.R. Barrenechea, F. Brezzi, A. Cangiani, E.H. Georgoulis (eds.), Building Bridges: Connections and Challenges in Modern Approaches to Numerical Partial Differential Equations. 115. M. Griebel, M.A. Schweitzer (eds.), Meshfree Methods for Partial Differential Equations VIII. 116. C.-O. Lee, X.-C. Cai, D.E. Keyes, H.H. Kim, A. Klawonn, E.-J. Park, O.B. Widlund (eds.), Domain Decomposition Methods in Science and Engineering XXIII. 117. T. Sakurai, S.-L. Zhang, T. Imamura, Y. Yamamoto, Y. Kuramashi, T. Hoshi (eds.), Eigenvalue Problems: Algorithms, Software and Applications in Petascale Computing. EPASA 2015, Tsukuba, Japan, September 2015. 118. T. Richter (ed.), Fluid-structure Interactions. Models, Analysis and Finite Elements. 119. M.L. Bittencourt, N.A. Dumont, J.S. Hesthaven (eds.), Spectral and High Order Methods for Partial Differential Equations ICOSAHOM2016. Selected Papers from the ICOSAHOM Conference, June 27-July 1, 2016, Rio de Janeiro, Brazil. 120. Z. Huang, M. Stynes, Z. Zhang (eds.), Boundary and Interior Layers, Computational and Asymptotic Methods BAIL 2016. 121. S.P.A. Bordas, E.N. Burman, M.G. Larson, M.A. Olshanskii (eds.), Geometrically Unfitted Finite Element Methods and Applications. Proceedings of the UCL Workshop 2016.

122. A. Gerisch, R. Penta, J. Lang (eds.), Multiscale Models in Mechano and Tumor Biology. Modeling, Homogenization, and Applications. 123. J. Garcke, D. Pflüger, C.G. Webster, G. Zhang (eds.), Sparse Grids and Applications - Miami 2016. 124. M. Schäfer, M. Behr, M. Mehl, B. Wohlmuth (eds.), Recent Advances in Computational Engineering. Proceedings of the 4th International Conference on Computational Engineering (ICCE 2017) in Darmstadt. 125. P.E. Bjørstad, S.C. Brenner, L. Halpern, R. Kornhuber, H.H. Kim, T. Rahman, O.B. Widlund (eds.), Domain Decomposition Methods in Science and Engineering XXIV. 24th International Conference on Domain Decomposition Methods, Svalbard, Norway, February 6–10, 2017. 126. F.A. Radu, K. Kumar, I. Berre, J.M. Nordbotten, I.S. Pop (eds.), Numerical Mathematics and Advanced Applications – ENUMATH 2017. 127. X. Roca, A. Loseille (eds.), 27th International Meshing Roundtable. 128. Th. Apel, U. Langer, A. Meyer, O. Steinbach (eds.), Advanced Finite Element Methods with Applications. Selected Papers from the 30th Chemnitz Finite Element Symposium 2017. 129. M. Griebel, M.A. Schweitzer (eds.), Meshfree Methods for Partial Differential Equations IX. 130. S. Weißer, BEM-based Finite Element Approaches on Polytopal Meshes. 131. V.A. Garanzha, L. Kamenski, H. Si (eds.), Numerical Geometry, Grid Generation and Scientific Computing. Proceedings of the 9th International Conference, NUMGRIG2018/Voronoi 150, Celebrating the 150th Anniversary of G. F. Voronoi, Moscow, Russia, December 2018. For further information on these books please have a look at our mathematics catalogue at the following URL: www.springer.com/series/3527

Monographs in Computational Science and Engineering 1. J. Sundnes, G.T. Lines, X. Cai, B.F. Nielsen, K.-A. Mardal, A. Tveito, Computing the Electrical Activity in the Heart. For further information on this book, please have a look at our mathematics catalogue at the following URL: www.springer.com/series/7417

Texts in Computational Science and Engineering 1. H. P. Langtangen, Computational Partial Differential Equations. Numerical Methods and Diffpack Programming. 2nd Edition 2. A. Quarteroni, F. Saleri, P. Gervasio, Scientific Computing with MATLAB and Octave. 4th Edition 3. H. P. Langtangen, Python Scripting for Computational Science. 3rd Edition 4. H. Gardner, G. Manduchi, Design Patterns for e-Science. 5. M. Griebel, S. Knapek, G. Zumbusch, Numerical Simulation in Molecular Dynamics. 6. H. P. Langtangen, A Primer on Scientific Programming with Python. 5th Edition 7. A. Tveito, H. P. Langtangen, B. F. Nielsen, X. Cai, Elements of Scientific Computing. 8. B. Gustafsson, Fundamentals of Scientific Computing. 9. M. Bader, Space-Filling Curves. 10. M. Larson, F. Bengzon, The Finite Element Method: Theory, Implementation and Applications. 11. W. Gander, M. Gander, F. Kwok, Scientific Computing: An Introduction using Maple and MATLAB. 12. P. Deuflhard, S. Röblitz, A Guide to Numerical Modelling in Systems Biology. 13. M. H. Holmes, Introduction to Scientific Computing and Data Analysis. 14. S. Linge, H. P. Langtangen, Programming for Computations - A Gentle Introduction to Numerical Simulations with MATLAB /Octave. 15. S. Linge, H. P. Langtangen, Programming for Computations - A Gentle Introduction to Numerical Simulations with Python. 16. H.P. Langtangen, S. Linge, Finite Difference Computing with PDEs - A Modern Software Approach. 17. B. Gustafsson, Scientific Computing from a Historical Perspective. 18. J. A. Trangenstein, Scientific Computing. Volume I - Linear and Nonlinear Equations.

19. J. A. Trangenstein, Scientific Computing. Volume II - Eigenvalues and Optimization. 20. J. A. Trangenstein, Scientific Computing. Volume III - Approximation and Integration. For further information on these books please have a look at our mathematics catalogue at the following URL: www.springer.com/series/5151